(How) do computer maps make money?
I keep coming back to something Tom said to me offhandedly when I was working on the original grant proposal for this dissertation a couple of years ago, which was that “maps are a loss leader.” Consider the following:
- Outside of satellite imagery companies, big incumbent navigation tech firms (Garmin, TomTom) and monolith tech companies with mapping divisions (Apple, Google, Microsoft), there aren’t really publicly traded geospatial software companies. (That’s not to say it’s never happened; MapInfo was a publicly traded company that had an IPO in 1994, but it got acquired by Pitney Bowes in 2007, and Navteq was publicly traded 2004-2008.)
- For the monolith tech companies with mapping divisions, it’s not clear whether the maps are a major source of revenue. For Google, Maps are probably some of their revenue, but it’s still probably nowhere near Search (personal recollections by past Googlers suggest that monetization of maps has not been easy; I have Thoughts on this that I should get down later)
- The industry incumbent, Esri, has been privately held since its founding, seemingly hasn’t ever taken outside investment, and seems likely to stay that way.
- Of the geospatial startups that emerged in the 2000s and 2010s, a very small number have been acquired by bigger firms seeking in-house mapping while the remaining ones are still relying on venture capital (which makes it seem sort of unlikely they’re profitable)
This isn’t to say maps and/or geospatial make no money. Forbes has estimated Jack Dangermond, the CEO and founder of Esri’s net worth for years and they’re pretty sure he’s a billionaire. It’s more that it seems like the market for geospatial software is smaller than the seeming ubiquity of maps and geospatial data would suggest, and it’s not a market that can do the kind of 10x scaling that venture capital tends to expect. The actual work of making and doing stuff with maps is niche; your market demographic simply isn’t going to be “every single person on earth” the way it is for Facebook or Uber or Airbnb.
One central source of suspicion of geospatial software companies by critical, lefty academics is that they are, well, companies and companies generally have a profit motive that supersedes any one worker or owner’s good intentions. That suspicion sometimes resembled outright hostility in the 1990s—not necessarily unwarranted hostility, but also not especially strategic hostility in terms of finding an opening to better understanding these companies and how they work. I’ve been wanting to work out a taxonomy of various geospatial business models for my dissertation because I think it’s helpful to be precise about the different structures of companies doing geospatial stuff and not lump all the companies into a “companies bad, boo!” pile. How do different business models and corporate structures affect the kind of technologies that come out of these companies? How do they stay afloat (or not)? What can that tell us about the moral economy of geospatial software?
The first thing that seems important to state upfront, even though it seems obvious: the business of maps is almost entirely business-to-business, not business-to-consumer. Even if a digital map or geospatial product is consumer-facing, most of the money changing hands doesn’t happen at the level of the individual looking at a map. This doesn’t mean there’s no company-to-individual transactions in these models: I personally have and pay for a Mapbox account to make API calls and run data visualizations. But I consider it a business expense because I use it for work. The only sector that does outright consumer products seems to be navigation hardware (e.g. companies like Garmin and TomTom—though notably both companies have B2B verticals). We could I guess also count print maps in here too as I’m guessing Rand McNally isn’t hand-drawing stuff anymore, but that’s “computer maps” on a real technicality.
Again, this seems sort of obvious, but I just haven’t seen much in the academic literature that lays this out or gets into the taxonomy of business models.
Software licensing
Maybe one reason the B2B-ness of it all isn’t totally legible in the GIS literature is that Esri’s public image is one of being very engaged with users: after all, their biggest annual conference is for them, not for third-party developers. But Esri software is not priced for consumer use (it’s also not actually very consumer-friendly considering it only runs on Windows machines), and its users are typically using the software in a niche institution—a university, a real estate company, a government agency, whatever. This is presumably a pretty lucrative niche if you look at the price tag for an ArcGIS license and imagine multiplying it by like, every state university with a geography or environmental sciences program.
Of course Esri isn’t the only vendor doing this, they’re just the one that most people know because they’re really big. I don’t feel like this one needs a ton of explaining.
Advertising
This is (in theory) the main way Google Maps makes money. You want people to know about your place, you pay money for it to be more visible at some zoom levels on maps.
Scaffolding for geospatial-adjacent businesses
The main market for this is developers for other companies: things like APIs and cloud tile hosting for companies that use but are not all about maps. DoorDash using Mapbox for routing and base maps is a pretty concrete example—and one that DoorDashers apparently have not been happy about, but as the linked Reddit post notes it’s likely that DoorDash is trying to save money not paying for the (likely more robust, but also probably more expensive) Google Maps API for routing instead. There’s more variety in kinds of customers you might have (newsrooms, white-label map widgets like product locators, dorks like me who make art using geospatial data); mostly you’re just trying to do sales. (ArcGIS’ services for publishing geospatial data online or doing GIS dashboard type stuff feels of a piece of this but I’m guessing that it is a facet of government licensing agreements.)
The limits of this model lie in the fact product scaffolding unfortuantely might enable a company to scale beyond you: at a certain point when a company gets big enough it might be cheaper to take the maps work in-house (Uber and Lyft did this, so did Facebook—which, once you’re running enough of your own data centers it kind of doesn’t make sense to be paying for tile servers somewhere else).
Moral critique of this business model sort of comes down to who a company is willing to take on as clients—which, in the early years of new wave geospatial startups, was definitely a point of pride for a lot of these companies but today has sort of faded to the background. Back in 2013 the CEO of Mapbox made a big show on Twitter of posting that the company wouldn’t work with human rights violating orgs or actors. Today, Palantir is a happy Mapbox customer. (Honestly, low-key surprised Palantir hasn’t just gone to in-house geospatial, but whatever.) I doubt that anyone at Mapbox is about to turn off Palantir’s API access.
Collecting and selling geospatial data (about land, not people)
This is technically just a sub-sector of geospatial scaffolding and I’m using to describe a business model that sort of doesn’t exist anymore, which is being a data collection middleman to companies who need maps: basically, the Tele Atlas and Navteq model. It’s pretty capital and labor intensive, especially when you consider maintenance (since places uh, change over time). Satellite imagery companies also fall into this category; they’re even more capital-intensive because they put stuff into space but they also have a weird problem of collecting so much data that almost no one ever sees. Sometimes I wonder how much it would cost to try and get only imagery of all the oceans on a single day just because people don’t ask for ocean imagery super-often and there’s something wild about millions of images that maybe no one will ever look at just sitting on servers somewhere because you can’t actually skip the oceans when you’re orbiting the earth.
Spatial data arbitrage: geospatial plus some other thing equals value
This coinage isn’t mine; it’s from a 2018 paper by Will B. Payne and David O’Sullivan about ZIp2, Elon Musk’s first startup which, in retrospect, it’s crazy that this man started his fascist empire with a product this banal. Musk was somehow able to get a free (???) copy of the Navteq database from one of the company’s founders (idk, they’re both South African?) and mashed it together with a dataset from a company called American Business Information to make software for generating like local city guides. This feels of a piece of data brokering more generally—linking seemingly disparate datasets to produce new value—and it’s less about one’s ability to do the capital-intensive work of data collection and more about having a good eye for a fun table join.
Geospatial and the surveillance economy: because that’s what you thought the previous section would be about, and maybe it should have been
So in some ways we could think of bundlers of consumer geospatial data as a slimier variant of the data collection middleman above: instead of sending around cars with sensors to groundtruth the world, they passively gather the locations of cell phones, aggregate it, and sell it. The reason I think that’s a little sloppy is that to do that, you don’t actually have to do any geospatial analysis or visualization. Data broker companies selling bundled location data like Near Intelligence’s Epstein Island visitor data aren’t rolling their own unique geospatial analysis stack. At best, they were probably querying PostGIS and based on the successor company’s website, it looks like they’re using the Google Maps API. They’re customers of geospatial software, not vendors of it.
It’s maybe petty, but I don’t like putting them in the same catetory as groundtruth data businesses because it feels like stolen valor: Tele Atlas data collectors drive around cities and calibrate sensors and run quality control checks and do stuff in the world; data brokers bundling location information have to like, make a shitty free app that slurps up data and/or know other data brokers to buy from. Put another way, there isn’t really a huge technical difference between the work of collecting location data from a phone and the work of collecting any other kind of data from a phone. Certainly the location data has worse broader ramifications for regular people, but I’m pretty sure it’s not de facto harder to obtain than an accelerometer or battery level.
Consumer Navigation
So in addition to selling devices to individual consumers, I think this also bleeds into work with car companies to build out onboard vehicle navigation tech. It’s a sector I’m honestly less attentive to because I’m not that outdoorsy (I like hiking and all, but I live in Brooklyn, man).
Niche Industry Navigation/“Fleet management” (basically, trucker surveillance)
There’s a scenario where I just did a dissertation about the history of this sector because I just think it’s so fascinating and under-studied. Truckers can’t go on all of the roads—sometimes because of low-lying overpasses, sometimes because of weight restrictions, sometimes because of hazardous material restrictions. Routing in general is an annoying and hard technical problem; routing for trucks is extra annoying. It’s made more annoying by the fact that for a certain (admittedly dying) generation of truckers, having to use a GPS device at all is sort of insulting.
Pre-GPS, there’s a few routing-specific businesses that pop up in the late 1970s and pick up steam (maybe not surprisingly) after deregulation of the trucking industry in 1980. When I talked to a former ETAK person about why they hadn’t gone for the logistics market I was told that basically Rand McNally had it cornered. There’s also a wild thread of the history of CDMA to fleet management—one of Qualcomm’s first financially successful products, Omnitracs, was a pre-GPS vehicle tracking system using satellite technology that for various reasons got its engineers thinking about ways to do networked communications differently, leading to developing CDMA. So uh, thank a trucker for your phone (I mean, they also probably brought it to whatever warehouse it came from to you so be thankful on multiple levels).
Doing any of the above specifically for governments
This is sort of the obvious “how to have sustainable business” option for all of the above business models, and it’s the thing that makes a lot of sense when you play the “who needs maps for stuff” game. It’s wild looking at the histories of a lot of geospatial tech and seeing its ties to public domain government-funded software that then became the basis for products to the point that governments mainly just contract with the geospatial companies now.
I’ve been trying to unravel the history of the Digital Chart of the World project and that seems like a story mostly about trying to unify defense procurement through developing data standards? That almost as a secondary effect produced a global basemap.
Unsurprisingly, the biggest federal US purchaser of geospatial software and technology is the Department of Defense (who spent more on Esri products in FY 2024 than the next-highest spenders combined—which are, incidentally, Department of Agriculture, Department of the Interior, Commerce, Homeland Security, Department of Justice, and the Environmental Protection Agency). This is the thing that a lot of critical GIS discourse comes back to—revealing the nefarious apparatus of state violence propping up maps software writ large, etc. Some of it is just being jaded, but I want more out of this framing than a sort of told-you-so defiance. I want specifics. I want to know what it would look like to take DoD off Esri’s balance sheets and how fucked the company would or wouldn’t be without them (and without defense agencies elsewhere in the world).
What am I missing?
There’s definitely stuff I’m not thinking of in here or different ways of thinking about it, and eventually I will have to revise this to sound more academic and serious. Happy to receive suggestions of other angles to look at.