Most of us open our map apps every day, but have you ever stopped to think about how they actually work, or how little they’ve changed in a decade?
In this episode, Jianggan and Shaolin (co-founder of NextBillion.ai) dive deep into the invisible world of mapping technology, the infrastructure powering everything from ride-hailing to food delivery, and explore why keeping maps accurate is one of tech’s most expensive and underappreciated challenges.
Tune in to discover:
- Why maps haven’t really evolved in 10 years — and why that’s a problem;
- The hidden cost of keeping maps alive (and how they “decay” without constant updates);
- How local players like Grab and Naver build their own maps;
- The role of AI in fixing global address chaos
Timestamps:
05:30:00 – Why maps stop improving once they’re “good enough”
10:00:00 – Local vs global: who should build the world’s maps?
15:30:00 – AI, addresses, and the promise of smarter navigation
22:00:00 – The future: AR maps and new ways of seeing the world
Featured materials:
Inside Luckin Coffee: How they launch 119 drinks a year | Impulso E131
WTF is Carbon Accounting?! | Impulso E89
E33: [Unfiltered CTO talks] US vs China team culture, business design, loneliness, the future
Also available on Spotify
[AI-generated transcript]
[00:00:00] Shaolin: You still fail to find a place, but the worst part today is that if you cannot find it on a map, it’s actually does not exist because how you can arrive there I remember they said perception is reality.
[00:00:13] Jianggan: Hello. This is episode 1, 3 2 of the Impulso podcast, and today we are having Shaolin again. So he has been on two previous episodes of this podcast. Before the first episode, he was talking about chewing glass, right? I mean, the journey of entrepreneurship. The second episode that he was in, he was talking about carbon counting, which was a whole sort of sophisticated subject on its own.
But today I think we are going to talk about something which is a little bit lighter, which is maps, right? So, Shaolin has been building maps for how many years? 10 years.
[00:00:51] Shaolin: yeah, this year is exactly 10 years.
[00:00:54] Jianggan: 10 years. So you’ve been building maps since 2015. Maps are very important because it sort of underpins of a lot of things we do.
I mean, many of you use food delivery. Many of you use e-commerce. You need map to see where you are so that the riders and the delivery people can find you. We’ll all use Google Maps to go to where we want to go, right? And or ways for people in Malaysia who drive, et cetera, et cetera. So I can’t imagine that that we can live the modern light on a phone without a map.
[00:01:22] Shaolin: True.
[00:01:23] Jianggan: Yeah.
So today, if you don’t have a map on your phone, you find it useless. So when Google disabled the Google framework on Huawei, the very first thing they need to build is actually a map, because without a map, mobile is not like a mobile.
So I do remember when we talked a lot about Luckin Coffee on this podcast. When they first collaborate with Maotai the Chinese liquor . The first day they sold like 5.4 million cups and I was told that the system crashed on a day because the map crashed.
[00:01:59] Shaolin: Oh, I can imagine that.
[00:02:00] Jianggan: ‘Cause a lot of people are looking
for it. If I want to buy this, and to find where the next nearest store is. And that loads the app API, right?
[00:02:08] Shaolin: Yeah.
[00:02:08] Jianggan: So how have these Map apps evolved over the last 10 years, or have they evolved at all?
[00:02:17] Shaolin: So, to be honest, right. So I hope if I say this, I won’t receive any letters from any lawyers, but I don’t think it has been evolved a lot. So the main experience, which is for you to search and then to navigate you to the place, I don’t think they have been actively evolving.
I still see the same challenge. Again, again, again, you’re still trying to search someplace, but the outcome is not accurate. You still fail to find a place, but the worst part today is that if you cannot find it on a map, it’s actually does not exist because how you can arrive there I remember they said perception is reality.
[00:02:58] Jianggan: So, the pain points that I have, maybe I’m not sure how representative it is on a couple of things. Then. First is that whenever I drive to a place, and if I use the map to identify that place, you know that in, in cities in Asia sometimes have very big buildings.
The map always leads you to the. The front of the building and it never has the point of interest as the car park entrance, which you can actually go in. So that quite often lists for lots of waste of time because I mean, you drive to another driveway, you have to do a big round to come back.
And the second I find interesting. So a lot of times the road are closed for certain reasons. Sometimes it’s accurate, sometimes not accurate. I remember like many years ago I had a discussion with people building a map from what the government’s here, and they were saying that they were talking to Google, but Google was saying that, okay, we should not accept the information from you because the point of interest is crowdsourced.
[00:03:51] Shaolin: Yeah, that’s correct. I think that’s also one of the major challenge. So especially those huge global map companies facing, so they need to be managing the map for so many cities. Hundreds of thousands of cities, small countries, et cetera. So,
[00:04:09] Jianggan: So historically we see that they will send out maybe through third party contractors like this, kind of like cars or vans with all the cameras mounted to capture the information.
And some of them will become images on the map application. Is that something that they have to do continuously or,
[00:04:26] Shaolin: yes, exactly. So that is the fundamental challenge about the mapping industry. So if they can deploy these kind of vehicles globally in every city, and they hire someone to drive that vehicle and collect for the map average.
At least every quarter. Okay. They will have a very good quality of map, but it’s about ROI. So if you are investing in something like that, the amount of investment is actually huge. So if you look at those vehicles it’s not. Simple vehicle. So those devices they deploy is, they’re actually much more expensive than the normal camera.
We are having, they also have LiDAR, all different devices and also considering the regulation. So you cannot freely map in a country in the world, a lot of places you need to apply for license, et cetera. So now you imagine you invest like that.
And you deploy hundreds of thousands of those vehicles you spend that amount of money and it’s nonstop money.
[00:05:29] Jianggan: So you have to keep like your
[00:05:30] Shaolin: Yeah. You keep doing like that. So one thing I would always like to share about mapping is that the mapping industry is not an industry that you can be the friend of the time.
Actually time is so unfriendly to you. So imagine you have been invest so much into a map for a decade and then you stop investing,
[00:05:51] Jianggan: then the map will decay.
So is that the reason why as we mentioned in the beginning, right? The maps have not evolved much because, I mean, maybe five years ago you already reached a stage that is good enough, like any additional feature, additional accuracy and stuff would from our point of view or not make sense, right?
You would spend a lot of money to make it like marginally better than what it is today.
[00:06:13] Shaolin: Yeah. I believe that is the fundamental challenge. So, because if you look at two decades ago at that time, the personal navigation, it’s not an app, it’s still a device.
[00:06:25] Jianggan: Yeah, I remember that.
[00:06:26] Shaolin: It’s a device like this. You need to spend like hundreds of dollars to buy that device and you need to pay extra for all the update. And then later we have those navigation. Apps or say map apps. At that time, it was not free. You need to pay like $10 to purchase that app, and you still need to pay subscription fee.
[00:06:46] Jianggan: Yeah.
[00:06:46] Shaolin: But then everything become free and those huge company, they just make their, their maps for free. And then it leads to this situation it’s been challenging to find a good revenue out of that.
And then the investment is also huge. Then we often see a map quality is not that great.
[00:07:05] Jianggan: I do travel to China a lot. And you live partially in Beijing, right? So when I go to Chinas, I see the maps there. They, have built a lot of features that we don’t see anywhere else.
For instance you can pretty much drive without looking at the map. So it will tell you, okay, which lane to do and where should turn, what should pay attention to.
So the, the voice prompt is good enough and quite often, I think in the last two years, what I realized that you can. Especially in some of the medium-sized and bigger cities pretty accurately predict the red lights, right?
[00:07:34] Shaolin: Mm-hmm.
[00:07:34] Jianggan: So what’s the justification for that kind of investment? Is that because it’s a big market?
[00:07:38] Shaolin: I believe it’s quite. Typical China. Right. So I remember when the Chinese big map company. They first introduced the celebrity voice instruction.
Then becomes like a race. I sign these celebrities and then I sign with the other. Everyone is trying to sign exclusive contract. You can imagine you need to invest a lot to sign those celebrity. So this, you don’t see anywhere else. Also part of the reason is those big maps company, they’re not independently operating. They belongs to those huge company.
They believe that maps is part of the infrastructure. They don’t need them to be profitable. So they just invest, invest and try to get as more user as possible. So, and then that leads to this kind of competition.
[00:08:25] Jianggan: I think it’s the same everywhere else, right? So the large maps nowadays are owned by large tech companies because they need these maps anyway to power their other services.
And part of that is offered as a free consumer sort of offering so that consumers can also help them improve and maybe get them some data about people movement and stuff. Yeah, so, so there’s some strategy value in that.
[00:08:47] Shaolin: Yes especially, I believe for search, right? So a lot of the search itself is related to some kind of map, maybe not.
A map, but can be related to a restaurant, to some place that is interesting. So I remember some data points like 20 to 25% of the searches somehow related to map. So I think it’s makes sense that for search engine company they own their own map infrastructure.
[00:09:13] Jianggan: Yeah. Because you certainly don’t want somebody else to control that part. Right? They can, bypass you. In many places in Asia we also see their efforts, right? For instance, in Korea. So you have a locally developed maps, which are good enough for the market and so the business model is the same, right? So you have the local tech conglomerate, which will spend money, invest, and build that. And Grab is building that for Southeast Asia. That’s also because I mean, it has lots of locations that these drivers needs to go to.
So you, you were part of that process, right?
[00:09:40] Shaolin: Yes.
[00:09:40] Jianggan: Do you think sort of across the world it makes sense for people to build local maps or does a calculation based on the market size?
[00:09:46] Shaolin: I believe it makes sense. Because maps is very a hyperlocal thing.
So it will be very hard for you to have some kind of unified model to map all the reality in the world. So take the example of South Korea, right? So it’s not a large place, the total amount of kilometers. You need to map. It’s limited so you know the upper limit type of investment you need.
But the challenge would be, for example, the language when I’m traveling there I see a restaurant nearby looks good, but all the characters is in Korean. Yeah. How can I input that into my phone? Because I want to search what type of restaurant is that. It cannot. And then you imagine even you deploy that vehicle, right?
The vehicle scan the whole street. You collect all these names, and if you are not local native speaker, very hard for you to. Understand. So imagine if you are a global mapping company, you sit in San Francisco, you try to do this, it will be very challenging and much more expensive compared to if you have a local company who is willing to invest.
So I believe that is the right business model. But still in the long run, we still need global map in some way. So, actually there are already some interesting effort that has been driven by some of the main company in the industry. So, for example, we do have a project called Overture Maps.
That is a more advanced version of OpenStreetMap. Okay. Which is the very first open source map. The idea is just exactly like I said, so for example, I am the main map provider in, in South Korea. So I have been mapping this region in a very good way. And then you have been doing this for Japan.
[00:11:30] Jianggan: Okay.
[00:11:31] Shaolin: So how can we help each other? So they establish this idea that each part to contribute their own data in some open source way. Then we have a new version of the global mapping.
[00:11:41] Jianggan: It’s kind of like a little bit more decentralized compared to current way that you have one company which does everything. Okay. And also on top of the map that there are lots of things which can be built. Right. I mean, you left the job at that large tech company to build this company together with a few friends called NextBillion.ai.
[00:11:58] Shaolin: Yes.
[00:11:58] Jianggan: And you build applications on top of the map. So tell us a little bit more about what you have been building.
[00:12:05] Shaolin: Yeah, so actually we are. Not looking at the consumer based market. Because that’s already too many companies there and the type of investment is too large. For a company like us, a small startup.
So what we have decided to do is we try to solve a different bunch of problem. Mostly for logistics, ride Hailing and for delivery. You gave an example earlier you mentioned that you try to navigate to a building but the GPS location may be just at the center of the building.
[00:12:35] Jianggan: Yeah.
[00:12:35] Shaolin: But what you really need is the entrance of the parking.
[00:12:40] Jianggan: So for instance, in logistic company, they want to go to the loading bay instead of car park or entrance. Right?
[00:12:44] Shaolin: Yeah. It will be even more complicated. So imagine the food delivery market in Singapore. So someone is.
Doing cycling. Someone is walking, someone is driving, someone is motorcycling. What is the right place for you to park? And imagine if there’s three different park slots. Which one will be closest to the final location you need to pick up or deliver? These kind of information is not available on the traditional map because the traditional map is aiming to serve.
Individuals like us in a day-to-day basis. But the problem of a day-to-day basis is because every day you drive from your home to your company and then your company back to your home. After one month, you know everything.
[00:13:25] Jianggan: Yeah.
[00:13:25] Shaolin: So you no longer need the map app
[00:13:28] Jianggan: So you’re saying that the consumers are less adventurous than what we typically want to think?
[00:13:33] Shaolin: Yeah. Anyway.
[00:13:34] Jianggan: Yeah. So that’s why the consumer maps, they don’t actively collecting those information because the information first is not easy to collect.
You need to have a data loop. You need to involve the final delivery so that they can share the real ground truth. Like a feedback to the map. The consumer base map, they don’t have that, and also that is the key challenge that those logistic companies are facing.
So we see the business value there. So we build enterprise mapping solution platform for those company.
So previously without such platforms. The option for them is to build themselves. Or just use a commercial map and let the drivers on the fleet on ground navigate, right.
Or,
[00:14:15] Shaolin: So traditionally if you go back to 20 years ago, there’s no good maps.
[00:14:20] Jianggan: So that would be two years before iPhone was out. Then you needed, still
need to have the small devices.
[00:14:25] Shaolin: Yeah, you still use very traditional phone. Right? So at that time, everything is managed by human. So if you have a bunch of deliveries you need to make, it’s a human to decide the order, the route, and then it relies on the driver themself. So there’s no technology, no efficiency. And then nowadays it’s, like we try to optimize based on the. Either the real time traffic of now or the estimated traffic in the next like couple of hours. It’s more advanced than before.
[00:14:57] Jianggan: And nowadays do you see AI and agents potentially change the landscape here?
[00:15:02] Shaolin: Yeah we always want to introduce this concept because it’s very good for fundraising
[00:15:06] Jianggan: okay.
[00:15:08] Shaolin: We have been thinking about that for quite long. So
[00:15:10] Jianggan: yeah, the company name has AI right NextBillion.ai.
[00:15:13] Shaolin: But to be honest, it’s quite challenging. So I would see two parts of that. The first part is the fundamental algorithm that help us to say, estimate the traffic or plan a route or optimize for complex jobs.
Those, algorithms is actually pretty traditional. So as of today, we don’t see a good way that. The large language model based AI can solve that. They do have some traditional AI type of model that can help us to, for example, have a better estimation of the traffic but not the latest ai. The second part of that is actually about addresses. At that part, we see a clear advantage. So especially in, remote areas or say undeveloped areas. Where this address system is not that. Developed. And in that case, people will introduce more of those kind of descriptive language to describe your address because you imagine Singapore, right?
Everything have a six digits.
[00:16:16] Jianggan: Yeah. Every building has a code. It’s pretty easy to
[00:16:18] Shaolin: Yeah. To communicate. Right. But there is a building there. Yeah. And it’s no proper address to describe that. Yeah. And that what you can do, the only thing you can do is you go to that main shopping mall with this name, and then you go north.
You see that big tree?
[00:16:32] Jianggan: Three blocks on the right.
[00:16:34] Shaolin: So that part is traditional mapping technology cannot handle, but now we have this large language model. We see opportunity that these type of undeveloped address system can be improved.
But again, it becomes a problem of the ROI. So, because if you try to use large language model. On top of each of the single address search, it would be too expensive.
[00:16:57] Jianggan: I remember back in 2017, I was in the Middle East advising one of the ecommerce companies, and at that time across Middle East, that was the exact problem you’re talking about.
Right. So it’s like 70, 80% of the recipients, they don’t have a proper address. That you have to. First call them. If it’s a new customer, then they will describe to you how to get to their house. And they will input that into the system and they will send that sort of exactly as you said, cryptic language to the driver. And drivers will figure out, and maybe they will need a call a few times and once they figure out. They might add some more description. And then it becomes their knowledge then how to then, how to get there next time. But the problem is that if this driver leaves somebody else comes in Yeah.
You have to do the same process again.
So how do AI exactly solve this by some understanding of the, the language better or,
[00:17:43] Shaolin: yeah, so these problem we call geocoding. It’s a, it’s a term in the mapping industry, which is to say you provide an address, we translate it into a GPS location. So the traditional engine is the, those kind of natural language processing type of technology. Try to find the structure of that piece of address. What is the country name? State name? It’s very traditional. It’s based on some preexisting data sets.
[00:18:08] Jianggan: I see, so rule engine, basically.
[00:18:10] Shaolin: Yeah. Rule engine. It’s not smart and it cannot handle different languages. Okay. Well, but with the large language model today. Without even do one single thing, you put an address and ask him to. Try to tell you, which is the country, which is the state. Okay. It can do much better than the existing engine.
[00:18:28] Jianggan: I see, I see.
[00:18:29] Shaolin: And you can tell that if at least the minimum you can do is you can, you should tell if it is a real address, or it is a descriptive address, which may be more human effort is needed there. That’s is at least minimum you can do. And we’re also trying to see if we can use that as a base address, integrate the searching capability into the ai. And based on your description, can we. Try to identify the, the possible GPS location, at least if we can give a range, it would be much easier.
[00:19:01] Jianggan: And also now or less on like Alibaba’s Amap. So they have an application called Amap, which is the most popular map app in China. And now you have Google Map, right? You have reviews on both apps. So, it’s interesting now. It has become the default. But I see a lot of improvements that can be potentially made, right? Because now you search on the map about restaurants and it gives you some places but somehow I feel that there’s room to be improved.
I mean, now you have a system which is good enough. I mean, of course there’s room for improvement, but would you be able to justify the ROI? Is that also a discussion?
I think it’s quite different. China versus Amap versus Google Maps. Right?
China’s justification is that my competitors are all doing that, so I have to do it.
[00:19:49] Shaolin: So on Google Maps. I think at the beginning, it’s just natural.
So they have the feature of review. but it’s not something they are actively promoting but now everything feels like it’s very important to have on the map because for example I’m here today. If I need to find a place for lunch how can I execute that search? The search would be like the restaurant near this place Galaxis so these. Type of search. It’s like 80% of my main search for restaurant which means it has to be deeply linked. And eventually I need to look at the map to see that where is the restaurant, which is the direction I need to walk. I think it’s a good way. And also the Yelp type of business, it’s already quite good. So you can get you can get revenue from the advertisement.
[00:20:35] Jianggan: So they’re willing to pay for advertising. ’cause restaurants constantly need to acquire customers.
[00:20:39] Shaolin: Yeah. Now Google Maps have a large portion of the revenue coming from the advertisement. So when you are looking at the map, you should notice those square icon. Which is sponsored. You see more and more of that. Couple of years ago when I tried to talk to other people. It hard for me to even find a sponsor, but now you just score a lot of sponsor there.
[00:20:59] Jianggan: Yeah. But it’s, it’s kind of natural behavior.
Right. When I search for a restaurant, I want to see what is good enough? What others say about it. So from search to look at review. Yeah. That’s kind of natural behavior.
[00:21:08] Shaolin: Yeah. It’s very natural. So that’s why I think it’s a good feature. And also this feature will bring you revenue.
[00:21:15] Jianggan: And you say that in China it’s different.
[00:21:17] Shaolin: It’s because the main restaurant review is actually Dianping. which is currently owned by Meituan. But Meituan doesn’t own any consumer map. So the consumer map, it’s competitor to them.
[00:21:30] Jianggan: So, just for the context Dianping is the largest food and local services review app in China. And it’s gone far beyond Yelp, right? It has the levels sophistication in terms of locations, in terms of reviews, in terms of the ability, the penetration into the merchants . So, of course a very different ecosystem there. So moving forwards, I mean like 10 years from now how do you see that the people’s interaction with maps different from what is today?
[00:21:54] Shaolin: That is actually an interesting topic.
So, because one thing I realized after I enter this industry is that how people understand the map it’s very different. So someone just good at it? Someone is not good at it. So there will be some people, they look at the map.
They look around themself. They still cannot identify themselves. On the map. ‘Cause it’s the way of thinking is quite interesting, right? You see something on the map, which is on your, I don’t know, left side or right side of your brain because you need to understand the logic.
The reason I said that it has not been evolved. Because 10 years ago, even 50 years ago. You’re using paper map. But exactly the same, the same way that we illustrate a map the same way we draw a map. The same way that the map represents themself. And the same challenge. People are facing from time to time. So for example, I’m still find it quite challenging in the new airport. When I need to find the pickup location for ride hailing. All the apps, they will have a guidance. Image text. Image text.
[00:22:59] Jianggan: Yeah. Step by step.
[00:22:59] Shaolin: But do you find it extremely easy? I never think so, because the situation would change the photo may take two months ago. And then a restaurant has changed. And then you compare it, you cannot find the exact match . and then the description is always not that easy to understand. So I think the way that map has been presented to human and the way human is reading it need to be. Reform into something new. So, I don’t know, maybe AR based I believe that is the future. So these is something we need to. Make a big change.
[00:23:32] Jianggan: So the way humans interact with maps. And with the, with the new form factors, AR, VR, whatever. So there could be something interesting there. But exactly how we evolve, we don’t know.
[00:23:41] Shaolin: Yeah.
[00:23:42] Jianggan: It’ll be interesting to find out.
Thank you very much for tuning in and it has been interesting discussion about maps. And please do follow us and like, and share if you liked this content.
And we hope to see you next time.
The post How Maps Quietly Control the Modern Economy | Impulso E132 first appeared on The Low Down - Momentum Works.
