In some respects, we are all drowning in data and one of the big challenges going forward will be filtering the data we have so we can make sense of the world by removing the noise. The Travel Time Platform approaches location-based search form the perspective of the time it takes to reach a destination. So in the example of searching for the nearest movie theater, the user will not be presented with a result that shows the options based on a certain travel time limitation as opposed to movie theaters in a straight line distance. In this interview, Charlie Davies walks us thought the challenges of building a search engine based on travel time and what this might look like in the future.
A platform that takes video and creates 3D mapping layers based on that data. The video can be from a variety of different sensors, does not need to be vertically looking down on the geography and each 3D output is georeferenced!
You are more than welcome to reach out to me on social media, I would love to hear from you!
In Conversation
The Idea in a Traffic Jam
Daniel: Welcome to the podcast Charlie. For those listeners that haven’t met you before, you are the founder and CEO of a platform called the Travel Time Platform. This has a lot to do with geospatial, a lot to do with mapping, and I’m really looking forward to diving in and learning more about it. But before we get into that, could you give me a little bit of background about yourself — how did you get involved in the Travel Time Platform?
Charlie: It’s a pleasure to be here and thanks so much for having me. There are many stories around how I ended up doing what I’m doing, but essentially I was working in a company that was working heavily on local search. This idea that there are two sides of a coin, where you’ve got people like me and you in our everyday life using the web and mobile to find things in our real outdoor environment — so away from our laptops, away from our phones — trying to find the hotel we want to stay in, the property we want to live in, the job we want to work in. We were at this company trying to figure out better ways of doing it. That sounds like such a big mandate: here’s something, make it better. We were working on several different ideas, and sadly, although that company had some great people in it and it was a great idea, it didn’t work out.
Charlie: So I was left back in my parents’ loft with some kernels of ideas in my mind, and one of them was an idea that I had while in a traffic jam on the M4 — one of the beautiful motorways in the UK. There was lots of traffic and I was thinking, we’re obsessed by distance. Every site that we had looked at, to develop our concepts or wireframes, there were always two different methodologies to deciding what to show someone. One was a municipality — you know, like the area of London, or Islington or Camden, or a city like Southampton. Those places exist mainly because they were areas that we could tax people on, and that grew up in the sales model online for local search. And the other one was distance.
Charlie: And distance is just the most useless way of searching for a human. Stuck in this traffic jam I just thought, when I’m searching for something that I need to get to, or it needs to get to me, then travel time is actually the most important unit to measure that by. So very naively we decided that we would go and build what we now very imaginatively call the Travel Time Platform, in order to do exactly that. We have this platform that enables you and I to search the world around us, and the data around us, in exactly the same way that we use it — by travel time instead of distance.
Daniel: So you realized that accessibility to different areas, to different things, is not necessarily based on a straight line distance between where I am and where I’m trying to get to. That’s part of the problem you’re solving. I think the other part of the problem you’re solving is the recognition that there is so much data out there that we need some way of filtering it. So it feels like you saw these two things and slammed them together and came up with the Travel Time Platform. Would that be a fair assumption?
Charlie: As much as Red Bull have plowed into the advertising leviathans that they’ve become, I’ve never grown wings. I can’t fly in a straight line, and I’ve never met anyone else that could either. So we’re at the mercy of those transport networks around us. And when we’re trying to interpret data around us as people, there’s so much choice, there’s so much out there. If you do any location-based search online there are usually thousands, if not tens of thousands, if not hundreds of thousands of results — and how do you find out that the one that you’re looking at is the most relevant for you at that time?
Charlie: And it wasn’t just the amount of data out there in terms of what could be searched. When you put the element of travel time in there — doing a distance-based search, there are a number of different methodologies to do it, but it’s fairly straightforward, especially if you treat the world as a flat plane. To develop what we’ve done, we’ve had to actually amass huge amounts of transport data ourselves, in order to build it in such a way that people can plug into it. Say they’re a property company, or they’re just trying to find something for the weekend, or using one of our tools in something like QGIS — they can very easily say, how far can I get within half an hour by public transport, walking, driving, cycling, at different times of day, and then overlay that with whatever data they want.
Nearby Isn’t a Circle
Daniel: I’m really glad you mentioned the data side of things, because that’s obviously a huge piece of the puzzle and I want to talk about it later. But for the moment I want to stay with this idea of travel time and how it’s bringing more relevant search results to us. It feels like you’re putting a whole different spin on that first law of geography — that all things are related, but near things are more related. It feels like we could also add time to that: they also need to be related in time, and perhaps time here could be a reflection of accessibility and the amount of effort taken to move things from one place to another.
Charlie: It’s a great phrase, that near things are more important — well, what’s defined as near? If we’re in the middle of a flat field, near is going to be anywhere within, let’s say, a 50 meter radius around us, and then a mile, and then two miles, and you just keep on going. But humans have built this huge civilization across the globe, and that’s put a level of complexity around what near means. Near is so many different things now. If you take air travel, that’s reducing the amount of time that we can move around the globe. But closer to home — I live in central London, and near to me is so much more important with travel time, because I can get so much further north and south from where I live than I can east or west in the same amount of time.
Charlie: When we started with this idea it was actually to develop a consumer offering. We wanted to build a tool that we could use in our everyday lives to find stuff around us that was near and relevant. The problem with distance-based search, and defining nearby in miles or kilometers, is that you’re inevitably going to find lots of stuff inside of that radius, that circle, that you just can’t get to. From where I am right now there are certain parts of London that if I did a three mile search — which isn’t really that far — there’s going to be loads of stuff that I can’t get to within maybe 30 or 45 minutes.
Charlie: But the other value of what we do is not just removing the noise inside of a distance-based search — it’s all the opportunities that you would never think about. If you were standing at say King’s Cross station right now and you were saying, what could I do within 45 minutes? You could be 30 or 50 miles outside of London quite quickly and find yourself in an area like Milton Keynes if you went from Euston. Those islands that you find that would be outside of that original distance-based search are even more important — they’re more valuable to you, because they are areas that you just wouldn’t have considered. And to include them in the distance-based search, if you just increase that circle size, you’re just going to increase the amount of noise that’s there. So it’s fascinating to look at some of the different ways that urban areas can open themselves up, and the accessibility of them, when you look at them through the prism of time and not distance.
Travel Time Islands
Daniel: When I visit your website you have all these amazing images there of what you describe as travel time islands. For me they resemble the results of a viewshed analysis — they are islands, there’s oftentimes a huge polygon with holes in it, and polygons that exist outside the biggest one. It really opens your eyes to what is possible, and it shows these places that are very easily accessible, and the ones that are perhaps closer to you in distance are not necessarily so accessible. I think it’s a really interesting visual way of describing what the opportunities are in these particular areas.
Charlie: And it’s all about making a decision as well. If you’re near a peninsula or a river, you can’t just jump over the river. So something that could appear — even if you’re right there and looking at it in line of sight — it can appear very close. I could almost reach out and touch it, but it’s going to take me say 15 minutes to walk down the road, over the bridge, and then back over to the same place and look back on myself in the same position I was in. So it doesn’t even have to be that much of a travel time to make it more relevant and human to what we can actually do in the geography that we find ourselves in.
The Data Problem
Daniel: I think you’ve done a really amazing job of describing the problem and how you’re solving it. I’d like to move on now and talk a little bit about the data that’s involved, because I’m imagining this is a huge piece of the puzzle.
Charlie: One of our first challenges about developing what we’ve developed is that it’s always about speed. My background — I’m not a geographer, I’m not a GIS specialist, although my vocabulary is getting better by the day. I was focused on how can people use search in a more interesting way, so our focus was always on consumer sites, and it had to be quick, because if things are slow online people just tend not to use them. I remember doing one of our first big demos and it took about two minutes to do what we can now do in about 100 milliseconds. And beyond that there’s infrastructure and the teams around us to make sure that when you do a search in Sydney or San Francisco or Tokyo or Cape Town, it’s just as fast.
Charlie: But as we’ve developed this beyond London, beyond the UK, beyond Europe, it is data. There’s no way to do a good travel time search unless you have good data, and that’s been a big challenge for us, because we’ve had to go and source some of that data. There’s lots of open data, but the problem sometimes with open data is that there’s no such thing as a free lunch when it comes down to it. That data might be in different formats, it might be incorrect, it might be invalid, and there’s a lot of cleaning and processing that we have to do to that data, as well as test it, to make sure that when you put it on a map and do a route and then put it into our travel time platform, it returns results that you’d expect.
Charlie: One of the big challenges we’ve had is the public transport data. There is no source of truth in this world, as far as I’ve been able to find, that says these are the transport operators that are in this city, this country, this town, this area — so that you can feel confident that you’ve got all of the operators in that area. And it’s not just operators, it’s also sometimes private services. South Africa has a huge amount of private services that are almost used as public services, and amassing that data for us has been a big challenge. We’ve got really good at it, and we’ll get much better, but you can’t do what we do without having the best data possible.
Daniel: When I think about data that might be applicable to solving these kinds of problems, I for sure think about infrastructure, roads, buildings, bike lanes, and you mentioned public transport and other transport networks. What are some of the datasets that perhaps someone like me wouldn’t think of initially that are really important?
Charlie: One of the ones I’ve been working on recently is the combination of different modes. There have been lots of new services — I’m going to talk about London because that’s where I spend the majority of my time — but services like on-demand bikes. In the UK we started with what we called Boris bikes, bikes that were at different stations, but now you’re having a number of different services like Lime and Uber with the Uber Jump bikes around, really changing the interchangeable sort of intermodality. If I get off at a tube station, maybe I’m not just walking now, I can actually cycle up the road much faster. So understanding the likelihood of where the positioning of those bikes is going to be can really change how you look at a city. Because most of the time if you’re using metro services in an urban area you can’t take your bike or anything else on there, but now there are services that you can append in order to do that. For me that’s been a really interesting one.
Charlie: But there are so many datasets out there that you can add in, and you sort of get into the world of the law of diminishing returns. I’ve been asked so many things in so many different meetings — I got asked once for horse riding, whether we would do that, and I think we’ve moved on slightly from there. Weather conditions are really interesting, but because most of our stuff is about predictive modelling — what I’m going to do tomorrow or the next day or over the next year or two — it’s not so much there. Congestion data can be really interesting, because people will actually make a choice: if someone’s doing a daily commute they may actually take a slower route knowing that the quality of their journey would be better. So we could keep adding, and I think we’ll probably never stop adding different datasets to the model to make it more realistic, more closer to what we’re doing in our everyday lives. But the law of diminishing returns really comes in when you’ve got too much data to manage and the results that you’re displaying aren’t really that much better as a result of it.
Which Modes, and How Much Choice
Daniel: You talked about all the different modes of transport that were possible, and I understand that you need to take those into consideration. But as a user, how do I say, well, I’d like to walk, I’d like to take a bike from here to here? Do I get different results depending on how I am travelling?
Charlie: We make our platform available to analytics departments — people in business who are making decisions based on locations. So, where should I put my office, where should I open my retail store, should I open or close this hospital for population growth in the future? We embed those into tools like QGIS, ArcGIS, Alteryx. But then our biggest markets are in the consumer sites. In the UK we have lots of property sites, but our recent biggest launch was with a company called Totaljobs, and that enables you to search for a job from your front door that’s within a commutable distance — whether that’s one hour, an hour and a half, or 45 minutes, whatever you set.
Charlie: It’s really interesting actually, because the first iterations we did of this, we gave every option that you could have on our API. How much time would you like to walk to the bus stop or your train stop? How much time would you like to be on public transport for? How much time would you like to walk at the other end? What exact time of day would you like to arrive at the office? And it just became far too much. When you look at route description and how you define a route, how you plan a route, you can really get lost in the detail — and the detail is all those things I just described. But the value of what you want is to understand very quickly the maps and the data that’s being produced by this mapping engine.
Charlie: So on the Totaljobs site you actually don’t get an option at the moment of what mode of transport you want. It actively selects which mode of transport would be best for you to go to that specific job, and it gives you the travel time that you would use in order to get there. So it’s not just saying everything within an hour and a half — it’s giving you the exact time to each of those jobs as well. So it’s about understanding huge amounts of data, taking the guesswork out of that, presenting it back to the user when they see those lists or those maps, and for them to go, oh cool, I can make a decision much quicker without having to copy the address, put it into Google Maps, do an A to B route, remember it, go back into the platform, do another search, find another result. We completely remove that guesswork from a user trying to understand what’s in front of them.
Maps Should Help People Decide
Daniel: That’s the really amazing thing about maps — they can give you that overview really quickly. I think sometimes we forget as practitioners in this space that we’re not talking about data sheets. We’re using a map for a reason, because we’re trying to give a visual overview so people can make that decision, and we’ve done the hard work for them. We’ve filtered out the data and we’re showing them things that are relevant to them, and not overwhelming them with every possible option.
Charlie: I didn’t study maps before doing this, and I don’t pretend to be a GIS specialist or something, but I am just fascinated by the use of maps and how people understand them and how people use them. There are so many maps online that are there just to make someone feel like they’re in the right place, but they don’t fully understand what’s in front of them. I always use this example of taking things away from a map but you still feeling safe that you know what you’re looking at. London’s a great example, because we’ve got the River Thames that ebbs and flows in this wonderful snake-like format through the middle of the city, and if I start removing things but keep that, most people would say, oh, that’s London, that’s great.
Charlie: And when you’re looking at a screen and seeing all the results in front of you — you put markers on a map, and actually when you ask someone to really look at it, they know that it’s in London, and maybe they know it’s in north London, south London, west, whatever it is, but they actually really spend quite a lot of time trying to figure out what each result means to them. And that’s a bit of a problem if you’re trying to get someone to do something with the results that you’re showing them.
Charlie: I had this thing I used to call spinning pavements. Mobile mapping is just amazing — you can pull out your phone, find out exactly where you are, and then find out exactly where you want to get to, and you can see this blue dot on a map in front of you, it’s literally in the palm of your hands, and maybe the compass is working to a degree and it’s telling you you’re facing in one direction. But I used to call it spinning pavements, where you just have someone looking down at their phone, spinning on the pavement, trying to orientate themselves to where they are. And they’ve got the most detailed map in front of them — the buildings around them, the building name, the street name, even the direction of traffic, that it’s a one-way street. So much data, but the cognitive decision-making around knowing exactly where you are was just completely lost.
Charlie: I find that in the GIS world maps have to be so detailed — it’s this scientific methodology about recording the environment that we’re in so we can reproduce it in different scenarios. But when it comes down to the people that are using these things around us, they really don’t interpret all that data. They just take the bits that they need and get on with their everyday. In a way people are their own GIS expert in their own world. We all do our own routes each day, we’re those local experts on what we do. We know that if we leave five minutes earlier we’ll have a seat on the train, or ten minutes earlier and we won’t get caught at that pinch point in traffic. And as you know, we’ve been collecting all this data, but we can’t use all that local information, because that’s what people know and what people do. But I’m always amazed that if you actually take data away and you just leave what someone needs, they can make a quicker decision and derive value from it far quicker.
Charlie: I can talk about this stuff for ages, but like the London tube map — it’s a map that is close to reality but not attached to the physical reality. It’s relative to each of the stations, and when you look at it you can understand it. I never used to have a favourite map, but I guess I do now, and they would be those pilgrim maps. Pilgrim maps were used by people on their religious journeys to go and visit various religious locations, and it would say to you, go towards this big tree, and when you get to that tree find that big hill, and when you get to that big hill find the next big landmark and keep going. These maps aren’t maps as we would look at a map on a screen or on a piece of paper — they’re instructions to use the world around us. And that’s what I think maps should do: they should inform people and tell people the information around them so that they can better make a decision. And that’s the core of what we’re trying to do as well — remove all of that guesswork, just give enough information so that someone can actually make a decision with the data that you give them. It’s just fascinating how we’re looking at our screens all day with maps, and then you look outside your window and there’s people outside using the world without even thinking about it.
Modelling the Future: Crossrail
Daniel: When I first saw your site and started to read a little bit more about what you do, it occurred to me that there’s a whole bunch of use cases here. We talked briefly about some of the integrations you have to QGIS and ArcGIS and presumably many other mapping platforms, so obviously the geospatial specialists in the room will be able to see the potential there. But I could also imagine things like proximity marketing. I could imagine doing some kind of modelling of the world — these are your options here and now, but if we shifted the world, if we added new streets or new public transport, we would get different results. Do you do any of that kind of work as well?
Charlie: We do, and we have in the past. I mentioned earlier on that when we had our first iteration of what we do it was clunky and slow, but it showed value in the end result. We found ourselves in this GIS market where people were using our data, and inevitably we got asked the question of what would it look like in the future. There are a number of fantastic infrastructure developments across the world at the moment — I’m going to have to go back to London again — because of Crossrail, a huge development of a train line. We had lots of people ask us, can you model this, can you map this? And we looked at some of the initial mappings that had been done actually by Crossrail itself, and they were a line across a map that showed where the stations would be, and we thought we could do a bit better.
Charlie: So what we did is we took as much information as we could find at the time — that’s got better as we’ve got closer to Crossrail being a real thing and actually opening — understanding when the trains are going to leave, how fast the trains are going to be, when they’re going to arrive, and integrating the future into a model that someone can use right now. But they can use it in the tools that they’re already using, so they don’t have to import a model into their own workstation, process all the data, try and figure it out for themselves. They can just turn a switch and look and say, well, what’s London going to be like in 2025, or what’s the Paris Metro going to look like in 2030?
Charlie: That can be really fascinating to look at how the city would change, because it is fundamentally changing the city. It may not look like it, but it changes how you could use it, where you could live, where you could work, where you could meet your friends after work or do something on the weekend. And it’s also really interesting for analysts and GIS specialists, because they can advise clients and the people that they’re working with to make better decisions — about opening a retail location here because in the future it’s going to get a lot of footfall, or property developers saying, look, we can build housing here because in the future it’s going to be pretty accessible. Not just because it’s near the station, but it’s near a station that intersects with that new line, so there’d be a lot more throughput on that network at that time. We don’t allow people to edit the platform on the go — there are some toolings out there that do that — but if you want to see what London would look like with Crossrail, you can just use our data in a couple of seconds, actually less than a second, and analyze your data against it.
Daniel: That sounds like a really fascinating use case to me. I love the idea of being able to speculate based on how the city is going to change, how these physical changes we make to our infrastructure are really going to affect the city and ultimately the people that live in there, and their lives, and the amount of time they spend transporting themselves back and forth.
Charlie: Time’s a really interesting thing, because it’s not a commodity that you can offset between your days. We only have a certain amount of time per day, per week, per month, per year, per person, per lifetime. And it’s really important to understand that we’re not only using our current environment in a way that we can maximize the amount of time to do the things that we really care about, but in the future that we’re building transport networks and looking at things in a way that we can optimize what we’re doing. We’re not going to try and slow people down and take time away from them, but perhaps even that gift of giving time back — when you find that you actually have some spare time to do things. I wish I had a better way to describe it, but it’s not a commodity that you can trade, so it’s one of the most important, valuable things we’ve got.
Culture, Data and What’s Next
Daniel: I’ve just got a couple more questions. Obviously we talked a lot about London, and a little bit about Europe — is this available in other places? Is it available in the US, Canada, Asia? And I’m assuming when we start talking about different cultures we have different rules around data. Have there been any challenges you’ve run into with respect to that?
Charlie: Absolutely. A car is a car, a train is a train, a bike is a bike, and those things behave in the same way — a car travels down a road, a train goes along tracks, a bike is two wheels and someone sits upon it and moves themselves around. But there’s a huge cultural element about how people actually use that, and what we’ve tried to do is develop the human way of using these vehicles, not just understanding what the vehicle can do. So when you look at a drive time model, we take into account where someone could safely stop and get out of their car — because there’s no point just understanding that halfway down a motorway you can stop, because you can’t safely stop. If you’re just looking at a drive time analysis for a road and saying, hey look, this is half an hour, I’m just going to stop — well, that’s not something that’s going to be helpful to that individual. It might be helpful to understand if your electric vehicle battery’s going to run out and you want to see how far you can get using the maximum of it, but there are so many different ways.
Charlie: We had to add cycling and public transport access in the Netherlands, because obviously in the Netherlands — I don’t know if you’ve ever walked around Amsterdam, but bikes are king. If you’re not looking left, right, up and down you’re going to get run over by a bike. They rule the roads and they take priority. In other cities they don’t. In the US cars are so much more popular than in Europe with public transport, because of the historic investment of governments being differently aligned. There’s a huge amount of work that we do in trying to understand that. I wish I could say that I’ve been to every city where we have data and I’ve spent time with a notebook walking around going through the detail of how people are using it, but it’s a really important thing.
Charlie: Another small one is that a lot of the trains in the UK have this thing where the doors shut 30 seconds before departure. In other countries, as the train’s leaving the platform you can grab onto it and on you go. That just is something that would never happen in the UK but would happen in other countries. And although the data and the maths stays the same — we’re looking at when something leaves, when something arrives — understanding how human beings would interact with that is another level of complexity, but also another level of making it more interesting and useful.
Daniel: You’re obviously someone who’s spent a lot of time collecting data from all over the world. I’ve often heard rumors about different regions and some of the challenges around data — for example I’ve heard that in Asia sometimes things are a little bit offset so people can’t get the exact picture, and I know there are some challenges around addressing in India. Can you say a few words about that?
Charlie: I was fascinated to learn that all the lat-longs in China are incorrect, because they have this algorithm that changes them. So when you think you’re looking at the exact object on a map, it’s not really there. We can’t host data in China because it would have to be through a joint venture, and it’d be far too much effort for a company our size to do that. Other datasets — for example in the UK there’s a mandate where we’ve got to release all the public transport data, same in the Netherlands. But I mentioned South Africa earlier on, where there are loads of private services, or just someone driving their minibus route, and it’s used every day by thousands of people, but it’s not a documented route. It’s not part of a government service, it’s just how people are using the world around them and using all the different modes that are available to them. And trying to understand that en masse is complex. You can’t always completely solve it from data, so one of the things that we try and do is get feedback from our clients and our clients’ users and then try and ingest that back into what we do, and also update the datasets that we’re using — this constant circle of data going around.
Charlie: It is really tough. Like I said earlier, there’s no direct source of truth for all the public transport data in the world. We’ve had to go out and build our own lists and our own datasets around that. India and its train network is one of the most fascinating train networks in the world — I never thought I’d say something like that, but I’m almost like a digital train spotter now. Trying to understand how you take the data from that, how much data you need to make a service useful. We’re not looking at optimizing the network itself, we’re just looking at showing people that they can interpret that data, or use that dataset with other datasets in a meaningful way. And we’ll never finish — this will never be a finished project, we’ll just have to keep going.
Daniel: I can see so much potential in this. When you look out into the future, what are the things, or the use cases perhaps, that you’re most excited about?
Charlie: I think there is going to be even more data at our fingertips that we can use to interpret and then provide more useful services to people. 5G — and I know there are so many buzzwords, you get sort of buzzword-centred on these things — but IoT, and understanding that there’s congestion on roads or certain public transport networks are congested at certain times, air quality, should I avoid certain areas and how can we design a route around that if I’m cycling at various times of day. There’s going to be so much more data coming out, and obviously a huge amount of computing power which is becoming cheaper by the day and more powerful as well. When you overlay those, the amount of data that we can just allow people to make those better decisions with. But the complex side of that is going to be trying to figure out where we can find the value in that to provide the services that we’re doing. It’s something that sometimes keeps me going and thinking, what other data can we add, what’s the next dataset out there that’s going to be available to us, so that when people look at maps, when they look at locations, maybe we can help them make a decision faster, or a better decision for them.
Daniel: Charlie, I really want to thank you for taking the time to do this interview with me, for coming along and teaching us all a little something about travel time and its importance and the potential use cases of it. We’ve mentioned the name of the platform a few times so people will be able to search that and find it, but is there anywhere else they can go to follow up with you or get in touch?
Charlie: My email address is charlie@igeolise.com — feel free to email me directly if you’ve got some feedback on our platform or any more questions. And if you want to have a quick look at the app, we have an app that you can test our data on, that very simply is app.traveltime.me. If you punch that into your mobile browser or your web browser on your laptop or desktop, you can immediately start playing with our data and seeing what it does. And I just want to say, it’s been an absolute pleasure talking to you as well, and thanks so much for inviting me on here. I feel very honoured to be amongst lots of other people that you’ve interviewed.
Daniel: That’s really nice to hear, thanks so much.
