Just under a year ago (which seems like a long, long time ago right now!) I had the pleasure of being interviewed by Ed Freyfogle for the then-brand-new Geomob Podcast. We discussed a wide variety of topics, including the joys of spotting Thunderforest maps on Polish TV, the challenges of building a business based on OpenStreetMap, the impact of large companies in the OpenStreetMap mapping community, business models, my advice for those thinking of starting their own business, and more. When the podcast episode was released a few weeks later, we received lots of great listener feedback, so I hope you enjoy it too!

You can listen to the episode using the embedded player here, or on your favourite podcast app (look for episode 10). I’ve also included a cleaned-up transcript below.

Geomob Podcast - 10. Interview with Andy Allan of Thunderforest (transcript)

Ed: Welcome to the Geomob podcast, where we discuss geoinnovation in any and all forms, be it for fun or profit.

Ed: Welcome back everyone to episode nine of the Geomob podcast. I am looking forward to today’s conversation because we have a very long time Geomobster, Andy Allan, who was based in London and now lives in Poland. He’s one of the early members of the OpenStreetMap community in London and also he’s the founder of an online mapping service called Thunderforest, so lots of things to talk about. Andy, welcome to the podcast.

Andy: Hi, thanks for having me.

Ed: Our pleasure. Our pleasure man. So let’s dive right in. What is Thunderforest? What do you do there?

Andy: Thunderforest is a company that I own and run that provides map services for small businesses. If you want maps on your website or in your application, you’re a developer, you’re looking for some great maps, then you come to Thunderforest. And we specialize particularly in activity maps. So instead of generic background maps or… you come to us when you’re looking for maps for cyclists or for hiking or public transport or some kind of specialist map like that. That’s really what we focus on.

Ed: And the maps are made using OpenStreetMap, correct? That’s the…

Andy: Yeah, we use a couple of minor sources, but 90% of what you see on our maps all comes from OpenStreetMap, and it’s really OpenStreetMap that allows our business to exist. If it wasn’t for OpenStreetMap, if it wasn’t for all the interesting details that are there, we wouldn’t be able to make interesting maps. So if we decided to try and use a different supplier like Tom Tom or TeleAtlas, they’ll have roads. Maybe they’ve got a few foot paths, but that’s it. And it would be really limiting on the kind of thing that we can make. But with OpenStreetMap - yeah, the world is your oyster, there’s all kinds of really interesting niche things that you can pull out of the data and display on the maps. So it’s really key that it comes from from OpenStreetMap and I was involved in OpenStreetMap long before I started making maps, long before Thunderforest appeared. So it was really a project and a company that’s grown out of a hobby, rather than a business idea that was looking for data.

Ed: Well nevertheless it seems to be a thriving business. You’ve been doing it for quite a few years now and I’ve seen the maps in lots of places so congratulations to you.

Andy: Yeah, thanks. It’s been going for more than 10 years now, so it’s growing rapidly, still, which is great. And yeah, there’s all kinds of different places that you find Thunderforest maps cropping up. It’s really pleasing when you spot somebody on the tram, or on the underground, using an app, and you see over their shoulder, you can recognize the maps. And we even had recently where we were watching a TV show, a Polish language TV show and I had to pause and point out to my wife, look, that’s my map up on the screen.

Ed: That’s awesome man, congratulations. That must have been a deja vu type moment. That’s cool.

Andy: Yeah, it’s really weird. But it’s kinda awesome when you can recognize something and, I mean if you’re a big enough map geek, even if it’s just like blurry in the background, you can be like, no, I recognize this because of that, so yeah, good moment.

Ed: As someone who also has a business that’s based on OpenStreetMap primarily, OpenCage, our geocoding service. We get asked a lot about maps because customers will come to us and they’ll say, “Oh, do you also do maps or not”? And we we very often recommend Thunderforest because you have so many great designs. I mean, as you say, some of the activity maps like OpenCycleMap, but you also have some really just wacky ones like Spinal Map, the heavy metal style, which is a lot of fun to look at. So what is the process? How do you create these styles? How do you get the ideas for them? How much work is it to create a style? How does that happen?

Andy: So there’s a few different places that the ideas come from. The most important one is existing customers. When they’re asking for something, like, they like our maps but it doesn’t quite fit what they’re needing for their new application. So those ideas come in and they pile up and eventually when enough people have asked for one of them, then we can develop that. So we’ve got a couple of styles on the go at the moment, which are under development but really based on what customers are looking for. And then of course there’s just having fun. The technology that I’ve been working on in order to make it easy for me to host different map styles means we can just do things for fun. And it’s not like there’s a big infrastructure cost or anything like that. So I’ve worked with Richard Fairhurst who you know well, and many of your listeners will know, and gave him free reign and said, come up with some styles that I haven’t thought of. And that’s where Spinal Map came from, that’s where our Pioneer map, which is a sort of 1800s themed railway map came from as well. Those are great ideas from Richard that worked really well with our infrastructure. And so it means that when other customers come to us, we’ve got these kind of “use your imagination” styles. So I had a games development company who came, they were looking for a medieval themed map, and they could really see from the Spinal map and from Pioneer that it’s not just serious maps that you can make, you can really push the limits with our infrastructure.

Ed: Yeah, this is really a case where the medium of a podcast doesn’t do it justice, so I strongly recommend all the listeners take 10 minutes and go browse the site and we’ll make sure we get links in the show notes that people can check it out. There are some that are really, really fun, but also some that are very beautiful. I like your topographic maps. How much fine tuning has to go on or is the map ever done? Are you continually finding new things to tweak and add and isn’t that a risk that you are just endlessly fine tuning?

Andy: Yeah, so there’s two aspects to the fine tuning. One is, it doesn’t take that long to make a new map style and to get the kind of broad brush strokes. So that it shows the features you want to show; gets that kind of use case. But then there’s all kinds of minor details, because the world’s a big and complicated place. There’s so many different things. Cable car stations are one that I was working last week on for the Outdoors map. We’ve had cable cars for ages. Actually when you’re using the map you realize sometimes that cable car stations need to be shown. And so that kind of endlessly adding more details or finding new places on the map where it’s not clear. Or I look at one of my maps and think - I could do this better if I tweak this a little bit, use a different icon, move things around a bit. That process is endless. And then there’s the second challenge, which is the big challenge with OpenStreetMap, in that OpenStreetMap is always changing. The way that mappers want to map features and enter that data, the tags that they want to use. Sure, some of the main ones stick around, but when you start getting into the more interesting tags and interesting features, all of a sudden they can decide: hey, this is a new tag for this, or we’re going to map it in a different way. And so you do need to keep on top of those changes as well. And that’s a key bit of value add that businesses can make on top of OpenStreetMap, is keeping up with those kinds of changes.

Ed: Well, this is for us, one of the, and it must be similar in your business, one of the big challenges that many people underestimate is they think, “Oh, I just need to set it up”. But actually OpenStreetMap is kind of a living beast, right? And the data is changing at such a rapid rate and new things are constantly coming out and it can be a real operational challenge just to keep everything running.

Andy: It’s a challenge. It’s okay when this is what you’re concentrating on. Where I see it, is other businesses who aren’t really focused on OpenStreetMap, they maybe set something once, they expect it to remain the same for a long time and you can see their maps start to degrade. Or they just make mistakes. Simple mistakes like assuming that the highway tag is for roads, but it also includes loads of other stuff as well. So sometimes you can see a company that has set up their own OpenStreetMap stuff and it’s like, yeah, I can see where you were going with that, but you didn’t really know the details. But for folk who are running businesses that are based on OpenStreetMap, we’re immersed in this kind of stuff. So when the changes happen, we just roll it out and move on.

Ed: Yeah. Well, let’s continue on this theme because as someone who’s been in the OpenStreetMap community for a long time, over the last couple of years we’ve seen more and more big players kind of enter the community in different ways. I mean the great example is Facebook is now doing a lot in OpenStreetMap, not always with the smoothest relationship with the existing community. As someone who operates a small business, how do you see that dynamic and is that an opportunity for OpenStreetMap, a threat? Is it both? What’s your perspective?

Andy: Well from the small business perspective then it’s all good news because these big companies that are coming in, like your Facebooks, I know the Apple uses OpenStreetMap data as well. They’re not competing against my business. They’re off doing their own thing. What they do bring in is some expertise and to OpenStreetMap some conflict as well, as you say. But the more that OpenStreetMap gets used, the more places in the world get mapped, the better it is for everybody who’s using OpenStreetMap. Because it’s one of the things I’ve noticed in the last four or five years, is, the most common question from prospective customers has just changed. When I started this business it was always “Where is OpenStreetMap complete?”, “Is OpenStreetMap ever going to be complete?”, “What about Germany?”, “How does it work in this country or that country?” So those questions have ended now. Like, people just know OpenStreetMap, they know that it’s big, it’s used by big companies, it’s fine. And that’s been a real boost to small businesses working with OpenStreetMap. So some of that is just time passes and OpenStreetMap continues to be successful. Some of it is boosted by these big players coming in and helping out with some of the mapping.

Ed: Yeah. What’s your perspective on a lot of these big players? Again, Facebook being an example, a lot of the mapping now is, let’s say, assisted by technology, be that image analysis of satellite images or whatever. How do you view that as someone who got started in the very, very early days? I mean, I think you were one of the first handful of OpenStreetMap contributors weren’t you? Like, riding around on your bike and with your GPS device?

Andy: Yeah, it was more than a handful when I was there, but yeah, certainly that’s how I started, riding around with a GPS. We didn’t have aerial imagery, all that kind of stuff. So my view on getting the external help is, so long as it’s providing tools to help regular mappers, I’m all for it. If you’re trying to bypass regular mappers and stuff data straight into OpenStreetMap then no, I don’t like that approach. And actually yesterday somebody linked to a blog post that I had written 11 years ago on this topic. I looked back slightly carefully going, you know, what were my views like back then? But actually it turns out they haven’t changed much in 11 years and I was still on that same track. So, if you build a tool that integrates with iD or JOSM or any of the other OpenStreetMap editors and it suggests changes or helps with another background layer or some kind of way to enable mappers and to make individual contributors more powerful… Yeah, I’m all for that.

Ed: Let’s switch tacks a little bit Andy. Because, you know, since you left the UK, it’s been a while since we’ve had you at a Geomob. But when you were a regular attendee, one of your, you know, we would often have startups come, or people come with their ideas and you would often kind of lead the questioning around the business model and how will this ever become a business. So, tell us a little bit more about the business side of your business. What’s your pricing model? How does it, how does that work?

Andy: Yeah. So I always have fun at Geomob with that question. But it’s a question that I reserve only for the right opportunities. Cause there’s loads of stuff that we see at Geomob, which is like, “Hey, I’m having fun, I’ve done this cool thing, look at my cool thing”. And I never asked them, what’s the business model? Cause if it’s, if it’s a cool thing, if it doesn’t need a business model, that’s fine. But when people come in and they haven’t really thought through what their business model is and it’s like, “Hey, we’ve got a team of people, we’re doing this professionally, we’re planning on making a business out of it but we’re just going to give everything away for free and work out what the business model is later”. Then, that always rings a little alarm bells for me because I’ve been there, I’ve done that. I’ve worked for a big VC backed startup where we just give stuff away for free and hoped it all worked out in the end. And it didn’t all work out in the end. So, I don’t want to see people making those same mistakes. And that’s the key bit about my business model. So I run a small business. I don’t run a startup. I don’t have funding, it’s all bootstrapped. Which means I need real customers who are going to pay money every month to my company. So I solve a problem, they pay me money, fairly straightforward business model. No kind of crazy advertising, nickels on the dollar, or anything like that. Just provide a good service and they pay. And the specific business model is a request based model. So we’ve got multiple different APIs. The developers, our customers, who use our APIs - whenever they make request to our service, we count them up. And that’s how much you pay every month. So we have a few tiers to keep the billing nice and straightforward. And yeah, that’s what people sign up to.

Ed: But do you ever get push back on that? Because sometimes people say, “Oh well, you know, I heard OpenStreetMap was free. Why should I have to pay?” We occasionally do get people with that type of attitude and that, I try to explain, “Oh well we’re providing a service and yes the data is free, but you know, we still have to keep the servers running”. Do you ever get that and how have you seen that change over time?

Andy: I don’t get it that much. I have a free tier available and so a lot of people use the free tier. I don’t know, maybe there’s just something around the wording on the website that that means people expect that they’re going to have to pay for it. I do have people who quite often question the pricing because perhaps they don’t feel it’s value for money. And in every case when you dig into it, it turns out that their budget is about $10 or maybe $20 a month. Those customers, for a small business, those customers are just not worth chasing. Like, it’s great if somebody wants to come in and spend some money, you know, that’s no problem. But for small businesses, you have to aim for customers who are going to pay a decent amount of money because you don’t have an unlimited support team. You can’t answer a thousand support requests a day. So you don’t want a thousand customers each of whom are only giving you $5 a month.

Ed: Well, yeah, it’s perverse. It’s almost the less they pay, the more support they require, in our experience.

Andy: Yeah, and that’s my experience with people as well who question the value, is they tend to then have lots of other questions and want lots of help. The customers who come in and go “Love your pricing model. It seems very reasonable for what you’re offering, I signed up yesterday”, then those are the customers you want. Like, they’re happy, they can see the value in the support, and things like that. And they are not trading off on the, “Hey, maybe I could do it myself” or “I found these random guys who set up shop six months ago and they were offering stuff free”. So there’s two sides to it. And as a small business, you want to focus on the people who are seeing the value in what you do.

Ed: Yeah. Are you typically getting people who are coming to you because of the aesthetics of the map or because it’s OpenStreetMap? Cause I know you’re quite well known in the OpenStreetMap community. What’s the dynamic there?

Andy: Listen it’s a mixture of things and I don’t have the most sophisticated kind of funnel tracking to figure out where people are coming from. A lot of it is the reputation of the maps that we’ve got already. We’ve been around for many years so people have heard of OpenCycleMap, have heard of our Transport map, they’ve heard of Outdoors. And so they know who to come to for that. Also because our maps are often quite a big part of what they see on other apps. So if they’re setting up a new business and they look at other competitors or similar fields, they can see these maps and they go right, where does this come from? So that makes it easier for me than for people like you because it’s not so obvious who’s powering the geocoding or who’s powering the routing. For the maps, it’s such a distinctive visual element that you can be like, “Oh, that’s an interesting looking map. Where does this come from?”

Ed: Yeah, you’re absolutely right. That is one challenge we face and I guess, your product does advertise itself in that regard. So, any advice for anyone out there thinking of starting a business based on OpenStreetMap?

Andy: Yeah, definitely. This is something I can stand in the pub for hours and chat about! The key thing is you first need to know who your customer is and focus on “customer”. Who’s going to pay you money for this. Cause there’s a million interesting things I can think of doing for OpenStreetMap, especially things that would be really useful for mapping and the mapping community. But with the best will in the world, they’re not going to pay you £100 a month or $100 a month to do this. I read an article many years ago which was really clear on this and it’s something I always recommend people to look up. It’s an analogy about what size of customer you want to aim for. And the analogy is, are you hunting rabbits, deer or elephants. And elephants are big companies like FTSE 100s or companies… basically they’re too big for a small business to aim for, especially for your first few customers. Rabbits are like end users, individual users, like people who sign up for Facebook or consumer services and there’s millions of them but they’re small and they are hard to catch and they escape really easily. And then deer are perfect for hunting. You got reasonable amount of food, they are reasonably easy to catch. You don’t need any specialist tools to get them and that’s definitely my advice for businesses is to go for primary customers are other small businesses. They have so many advantages. The best one is if you’re talking to somebody in a small business about your product and you’re trying to get them, to convince them. If you’re talking to the person who owns that company, who built it from scratch, has all the decision making authority, then you’re done. When they say, “Yeah, I’m going to sign up”, then they just sign up. They put their card information in and that’s the end of it. If you’re going after big enterprises, you can spend months going through their procurement division, or especially one of my customers has a separate company that screens their suppliers for them, and that’s just a whole lot of hassle.

Ed: Yeah. I can attest precisely to what you’re saying. We have the challenge, and maybe it’s similar for you, is that very often we get kind of discovered by the software developers at a company and they start using it and they like it and then they say, okay, now the projects going forward, like it’s time to, you know, we need to increase the volume, we want to become a customer. And then at that point you’ve got to deal with the person who can actually make the decision and pay the money and Oh man, it can be a pain.

Andy: Definitely. So that’s a key bit of building a business in front of OpenStreetMap instead of just doing something cool with OpenStreetMap. So if you’re, if you’re clear on your customer. It also helps define what you want your product to be as well. It needs to be like, like you said, we’re both in the same industry. We’ve got something that’s useful for developers, but then sometimes they’re not people who choose to spend the money. And if you’ve got that clear in your head, then it really helps with things like website design and the messaging and the different audience that you’re talking to on different parts of the website. So like API documentation, it can be written quite distinctly from the, “Hey, sign up, we’re a …”, you know, answering the questions that the purchasing manager has.

Ed: Yeah. There’s a real art to it, to navigating that and getting the right tone on the right page, in front of the right audience.

Andy: Oh yeah. And this is again where being clear on what kind of company you are chasing can help. Because I’ve come across places where developers are using our services, the payment teams are coming in and starting to cause too much trouble, and sometimes just like, it’s fine to walk away from that and say, “I’m not betting my business on needing this one customer deal to succeed”. And it takes a lot of pressure off. If I only had three large businesses as my customers, then a whole lot is invested in trying to navigate their internal bureaucracy, and that can be a complete nightmare.

Ed: I agree. You’re preaching to the choir as someone who has had to navigate that very path. It can be a nightmare. All right. So what’s next for Thunderforest? What does the future hold for you?

Andy: Well, there’s a whole lot of interesting technical things, but I’ll leave that for offline. We’ve got some more map styles and some more APIs coming up. The big thing for Thunderforest is trying to maintain the edge on technology. So a lot of stuff is moving towards vector tiles, it’s not the solution for everything. But we have our Vector Tiles APIs available. We’re working on that. We’ve had two releases of new vector tilesets in the last 12 months.

Ed: Andy, before we go further, maybe just for the benefit of this audience, what are vector tiles? Very quickly break this down, just want to make sure everyone’s aware, cause it’s one of the key technological changes I think going on right now.

Andy: Yeah, sure. So most people will be more familiar with raster tiles or normal image tiles, which is when you look at a map, what gets sent to you over the internet is premade small square pictures and sometimes, especially if you’re on a slow connection, you can see those pictures popping up on your screen. They’re great, they fit a whole load of use cases. But there’s a few edge cases where having the images already made before you send them over the wire isn’t the best option. So vector tiles are a different way of achieving the same maps, but where you send the raw data over the internet and either your phone or your web browser colour in the maps for you. So that means you can do some interesting things involving having different map styles without having to fetch more data over the network. It can work well for offline use cases and slow connections where, depending on your application, sometimes it’s quicker to pull the raw data over and draw the maps locally, than it is to fetch the pre-rendered maps.

Ed: Okay. Excellent summary. So, you say you’re moving, everything’s moving much more towards the vector.

Andy: Yeah, so we’ve been using them behind the scenes for years, now. All of our raster maps that we provide are actually being created using vector tiles behind the scenes. But the problem with the vector tiles, or one of the main downsides, is they need a lot more CPU power, a lot more processing power on the devices to draw. And it just wasn’t feasible 10 years ago to be doing this. As everything gets more powerful, more and more people are doing them online. So we’re seeing more of our customers taking the vector tiles from us, and doing onboard rendering. So that’s where we’re focusing a lot of our development effort, on making this easier for people to get started with, whilst still keeping our our edge on the custom map styles and all the interesting things we’ve done before.

Ed: Well, sounds good. Congrats on how far you’ve come and I look forward to all the future map styles. I’m sure there’ll be cool. Hopefully, obviously you’re always welcome to come speak at a Geomob anytime to tell us about your progress.

Andy: Yeah, thank you.

Ed: Our traditional closing question, as we kind of wrap up here. Looking back as a long time attendee, any favourite Geomob talks that stood out for you?

Andy: I mean, this is a really hard question because I’ve been doing Geomob since the start. So, I’ve missed the last a year or two, but there’s hundreds, there must be hundreds of talks that I’ve been to. The one that immediately sticks out in my mind though was one a few years ago from Anna Powell-Smith who was talking about land ownership and property ownership in the UK, and using different data sets, different open data, in order to investigate how much property and how much land was being owned by offshore companies. I really liked that because immediately, I think even whilst she was still talking, I was on the website that she was talking about and looking at my local area and finding houses within a few hundred meters of where I live that were owned by shady offshore British Virgin Islands companies. And I thought that was really interesting.

Ed: Yeah, great. That was a great project and also a great presentation that she gave and she won the best speaker prize. But also, you know, I’ve seen it cited numerous times, in media and I think it’s become a tool that a lot of people are using to try to understand the situation better. So good choice.

Andy: Yeah, and I think it’s a great story around open data and open government data as well, which I think, it’s really important to have those stories. And to tell the stories well.

Ed: It’s interesting, you’re the second person actually to mention her. Steven Feldman also mentioned her talk in an early episode, so we’ll have to get her on the podcast here and get her to talk about it in more detail. OK Andy, what’s the best way for people to learn more about you and about Thunderforest? How can they get in touch?

Andy: For Thunderforest, thunderforest.com has all the details. If you’ve got any questions, there’s a link on that site. Just send us an email and we can answer your questions. For my personal stuff, I think Twitter is probably the best place to see what I’m looking at and what I’m thinking about it, and I’m gravitystorm on Twitter.

Ed: Excellent Andy, thanks for coming on the podcast.

Andy: You’re welcome.

Ed: Thanks everyone for joining us today and listening to the Geomob podcast. Hopefully you’ve enjoyed the discussion. Please don’t hesitate if you have any feedback for us or any suggestions for topics that we should cover in the future. You can get the show notes over on the website, which is at thegeomob.com. While you’re there, if you’re not yet on the mailing list, please do get on the mailing list where we once a month send out an email announcing future events, summarizing past events and just generally sharing events that you may find of interest. You can also of course follow us on Twitter where our handle is geomob. You can follow Steven at stephenfeldman. You can follow me at freyfogle. You can check out Mappery at mappery.org and of course if you need any geocoding, please check out my service, which is opencagedata.com. We look forward to you joining us again at a future episode, and of course seeing you at a future Geomob event. Hope to see you there soon. Bye.