The Commercial Satellite Imagery Business Model is Broken
Disclaimer: I work at Azavea, and while I’m writing about topics relevant to my day job, I’m purely speaking as an individual without my employer’s input or permission.
Have you ever tried to purchase a satellite image?
It’s closer to a hostage negotiation than a sales transaction. There will be email introductions, phone calls, lists of demands, let-me-get-back-to-you-on-that-s, appeals to lawyers, requests for paperwork, and ultimately…a ransom. Often, the captive is never released.
Over the past two decades, commercially available satellite imagery has evolved from an absurd aspiration to a mature industry. Millions of people now carry an instantly accessible, global, cloud-free, high-resolution mosaic of satellite images in their pocket. That kind of access to exquisite satellite imagery was exclusively reserved for government intelligence officers until recently — Google Earth wasn’t even launched until 2005, for goodness’ sake!¹ The success story of non-governmental satellite imagery is so preposterously unlikely that I struggle to comprehend it.
And yet…the experience of buying satellite imagery is still so shitty it feels intentionally awful, like satellite imagery companies don’t actually want you to buy their product. It’s a 21st century industry following a 20th century sales playbook. Over the years I’ve come to realize the problem is not personal. The sales people want to sell you the imagery. The engineers want you to use the imagery. The executives want you to tell inspiring stories about the imagery. But the company does not care about you. You are a misallocation of valuable resources, an unprofitable and noisy distraction from what really matters.
It’s really that simple at its core. For satellite imagery companies, there is an economic and structural disincentivize to sell imagery to anyone besides the government (except in rare, acute cases, but you can’t build a business around rare, acute cases). The more I’ve ruminated on it, the more I’ve begun to think that the promise of commercial uses for earth observation data is premised on a lie. I’ve seen the lie many times in many forms: pinned to the top of flashy startup websites (some of which I’ve helped to author), scrawled across slides at conference presentations, and buried in VC investment memos. The lie is this: someone, somewhere is making a ton of money using satellite imagery for something besides military intelligence gathering.
The Curse of the Consummate Customer
Maxar is the largest commercial satellite imagery vendor in the United States, and probably the largest imagery-focused company in the world. Like a twin absorbing all of the nutrients in the womb, Maxar (“DigitalGlobe” at the time) devoured its rival, GeoEye, in the Great Merger of 2012™ after the Federal government decided not to renew GeoEye’s major contract with the National Geospatial Intelligence Agency (NGA). After a decade of DoD-funded innovation following the events of 9/11, the burgeoning earth observation industry was reduced to one monopolistic vendor serving one monopolistic customer.
And to be fair, the US military is the greatest customer of all time— happy to pay more than anyone else can afford for the imagery, working on a genuinely important mission, content to lock in long and lucrative contracts, and desperate to fund R&D on top of that as well. They are a strategic investor that takes no equity and yet can’t afford to let your business fail. Not only do they make the market, they also regulate it — with a maniacal grin on their face, the government wields a carrot in one hand and a stick in the other. The trouble with striking upon the greatest customer ever is that there are no other customers like them, by definition. Who can compete with that?
If you drew a chart depicting the monetary potential of each of the multitudinous applications of satellite imagery, I strongly suspect the distribution would follow a “power law.” In fact, if you summed up the area under every non-defense use case, it likely wouldn’t even equal the area under defense. Agriculture, energy, real estate, logistics, the whole laundry list — they just can’t compete with defense. I don’t want to confuse “monetary potential” here with “value” — I believe the “value” of satellite imagery’s role in climate science alone matches or exceeds defense…but good luck capturing any of that value. Such is the beguiling nature of working in an industry that can do a lot of good for the world, but generally can’t afford to.
The Innovator’s Dilemma
In 2013, while DigitalGlobe and GeoEye were still in the honeymoon phase after their shotgun wedding, Planet Labs was launching its first satellite. They promised to be different. Planet spoke of a lofty “Mission 1” to image the entire landmass of the earth once a day, every day. They evoked a sense of wonder and opportunity — what would be possible if we could take a daily pulse of all human and geological activity?
Planet’s founders strike me as earnest, starry-eyed hippies with rocket science degrees and soldering iron burns on their hands. I once rode an elevator with Will Marshall, the CEO of Planet, and mentioned that I thought Planet had a unique role to play in supporting assessment of the Sustainable Development Goals (SDGs) set by the United Nations. He had already thought of that — in fact, he was able to list all 17 of them, in order, from memory. He didn’t strike me as a soulless capitalist seeking the highest return on equity for his shareholders. Rather, he struck me as a passionate and brilliant visionary who accomplished an incredible amount in very short order, and who clearly saw the potential of his life’s work to do good in the world.
Still, eventually, Planet succumbed to the siren’s song of the defense world. It started simply enough with a few announcements of DoD contracts — just a few million bucks here and there…a drop in the proverbial bucket. But things clarified a bit at the end of 2018 when Planet acquired Boundless, the beloved service provider built on top of a suite of open source geospatial analysis tools. Shit immediately hit the fan. A big chunk of Boundless’s staff were laid off, and six month later someone leaked to the press that Boundless’ acquisition price was cut by 60% due to a purported failure to disclose material information. In a nutshell: everybody lost money and lots of people lost their jobs. The strategic rationale behind this ill-fated acquisition was that Planet needed additional support for selling into the government, and Boundless provided existing contracts and experienced software and support staff. The remnants of Boundless today are manifest in Planet Federal, Inc., a newly formed subsidiary dedicated exclusively to the pursuit of government work.
Strictly speaking, the innovator’s dilemma refer’s to Clayton Christensen’s brilliant observation that incumbent businesses cede ground to upstarts because they enjoy a high-margin, mature customer base and the upstart focuses at the lower-margin, higher-risk periphery that simply doesn’t make sense for the incumbent to go after. Over time, the upstart gains traction and expands more directly into the incumbents territory — often with a scrappier team and a newer, flashier product. I think this applies to Maxar: why support a startup ecosystem that complains when asked to spend a fraction of your main customer’s hourly spend? It’s an irresponsible allocation of resources to support them. It takes a ton of work, and what’s the upside, really?
Here, though, I’m using the term with tongue firmly in cheek. Because the dilemma for the innovator, in this case Planet, is that in order to thrive (and perhaps even just to survive), they’re starting to look a lot more like the incumbent than they look to be disrupting them. Mission 1 gave way to what I’m referring to informally as “Mission 2,” which appears to be: launch more high resolution satellites and make money from tasking them just like Maxar does. The dilemma Planet faces is whether they can retain the soul of the company while doing what is necessary to please The Greatest Customer of All, which is the only pathway to sustainability anyone has figured out so far.
To the Satellite Companies: You Have a Choice
If you’re reading this and you work at a satellite provider: you have a cool job. You work on undeniably meaningful technology with profound influence on the way the world operates. If you don’t love your job, it’s almost certainly a failure of management or of imagination. You’re sitting on one of the greatest underutilized digital assets in history. The archival data from the WorldView satellites alone is capable of unlocking untold billions in value for climate science, humanitarian aid, navigation, urban planning, small landholder agriculture, public health, and a host of other unprofitable and fantastically important industries. If you aren’t advocating for making it easier for people to afford, access, and exploit your data, then it’s time to start.
Here’s the choice as I see it for Planet, Maxar, and the rest: do you want to take an active stance and work to grow your industry? Or do you want to milk your most significant customer for all they are worth? I genuinely don’t begrudge you for choosing the latter — you have a duty to your shareholders and to your employees to make money and stay in business. If you genuinely believe it’s a choice between surviving and dying, then survive. But I have a hunch that’s a false choice — I think you can actually please the government while still growing the industry, but you’re going to have to live with delayed gratification. And I mean delayed gratification; it’s a multi-decadal bet if you really believe in the power of satellite imagery to transform the commercial industries you talk about all the time. Because no one gives a shit about imagery in any of those industries. They care about timely, relevant information they can use to make better decisions. And the gap between imagery and what they need takes a long time to figure out and is usually not at all obvious. It’s not going to show up on your quarterly earnings reports, is what I’m saying.
But if you are genuinely committed to making a thriving commercial earth observation industry possible, there are three things you can change to drastically improve your chances of success:
By this, I mean the sales process I alluded to at the beginning of this essay. I don’t want to talk to sales people at all, let alone sales people who can’t answer any of my technical or legal questions. You don’t equip your sales people to sell to technical startups, nor do you incentivize them to, I suspect. So stop putting them in a position where they’re having to do me a favor. Remove them from the equation entirely. I should be able to hit an API or submit a request for exactly what I want without asking anyone for permission. How is it that banks have figured this out and you haven’t?
Distribution is not enough. It’s not even close to enough — perhaps more important is licensing. Stop treating your archival data like a goldmine. If it were a goldmine, you’d be making money from it. I understand your predicament: you sell your archival data to the Federal government for a boatload of money and no one else can afford to match the price they pay you. But you don’t have to give us the same licensing — it can be an inferior set of rights and therefore a dramatically different pricing arrangement.
If you give data away in an open data program, license it for commercial reuse. When you license it for non-commercial reuse you stifle all of the interesting innovation that can occur which would enable the communities you’re trying to help to actually get the expertise they need to exploit your data. It’s not “enabling” to restrictively license your “open” data. There is no true threat to your paid products by permissively licensing data you give away unless you start giving away several orders of magnitude more data. When you go to an ice cream shop, they let you taste the flavors you’re interested in. But they don’t make you promise to spit them back out once you’ve tried them.
“Internal use only” licensing is restrictive enough that it should be dirt cheap. Multi-use licensing with a restricted number of affiliates is a little better, but it’s not materially different than internal use only, it just widens the aperture of what’s considered “internal.” Create a standard pricing structure and license for use cases where the customer wants to turn around and give the imagery away or do whatever they want with it. Think of it like paying to openly license the data.² Just figure out a price and then lower it, because you aren’t making much from all that data gathering dust on on your cloud servers (which cost you millions of dollars each month to maintain anyway).
One last bit on licensing…
Allow me to disclaim: I am not a lawyer, and this is not legal advice.³ When publishing openly licensed data, the specific licenses I would recommend to make the endeavor as valuable and productive as possible are CC0 1.0 Universal (CC0 1.0)Public Domain Dedication or Attribution 4.0 International (CC BY 4.0). These are two of the most permissive and widely-used licenses you can choose from. A ton of work has gone into these licenses, and they’ve be litigated on quite a bit (or at least their predecessors have), which makes them relatively safe and familiar for those of us who want to use it but don’t want to run afoul of the law. Whatever you do, do not invent your own license. That would be like inventing your own new type of combustion engine — it might be a fun hobby but it’s no way to mass-produce a car.
Seeing these super-permissive licenses, you may be worried, “what if people use our data for terrible things and it blows back on us?” Well, I’m here to tell you: the people that use your data for nefarious things don’t check the license first to make sure it’s ok. You’ve already let the cat out of the bag with your existing open data, so you might as well go all the way. I believe the good to come of it will far outweigh the bad, but that’s a judgment I am not ultimately responsible for (morally or legally) so I don’t envy your position. If my ideal licenses are simply too permissive for your taste, then a share-alike license like CC BY-SA 4.0 is the next-best thing, but know that’s going to dissuade the majority of your potential commercial user base from incorporating the data into anything meaningful. Odd ducks like Azavea, where I work, are happy to work with share-alike licenses because we try to give away most of our code anyway. But we’re a services firm, not a product firm: we benefit most when the broader market grows (rather than maximizing our share of the market), and opening up data and software helps spur overall growth. But even we prefer the more permissive licenses, because let’s face it: information wants to be free.
Charge as much money as you can for tasking satellites. You can point a giant, floating telescope anywhere on earth with low latency, and with all the new satellite launches you have planned for the coming years, your capabilities in this area are only going to get more impressive. If someone really needs an image of some spot on earth right now or once a week or ten times a day or whatever, make ’em pay for it. It’s crazy you can even accommodate that.
But for everything else that’s leftover, stop being so gosh darn greedy about it. For the most part, you already got paid to acquire it (Planet, I’m not talking to you — obviously Planetscope does not apply but SkySat will). If you just gave 20% of your imagery away, how much would that spur innovation and grow the market? What would the true marginal cost above your current baseline be to make it happen? You’re already storing it, visualizing it, and marketing it. Lack of affordability of archival imagery is stifling innovation because there’s so much of it you can’t possibly look at all of it with human eyes. You need to train machine learning models to mine it, and that means iterating over large swaths of your data a bunch of times during expensive computations. Your potential customers are thinking, “I’m not sure this idea is going to work, but it would be super valuable if it did…how much can I really afford to test it out if there’s a high risk of failure?” If instead it was free to test ideas out and only cost money to apply those ideas in practice, suddenly the risk profile would be dramatically altered.
And stop trying to “rent” analytical products. You can rent out basemaps and visual products being used by humans as a visual reference. But if you try to rent the images themselves, you suddenly have to accommodate every conceivable computation and transformation every customer wants to do. None of you are cloud computing companies— Maxar made a big show of driving their archive in a big rig to upload it to AWS and Planet quietly traded ~15% of its equity to Google for Terra Bella (now called SkySat) and a sweet cloud hosting deal (I’m assuming). You can’t compete with AWS or GCP or Azure because you’re built on top of them, so you’ll always look like a raw deal compared to just owning the imagery outright and using private cloud accounts or working off a local machine.
This industry is so young, it would be ridiculous to claim it’s too set in its ways to change. I’m not a cynic. In fact, I’m more hopeful than ever about the potential of satellite imagery to positively impact the world. But the satellite imagery companies need to meet the startups and research groups clamoring for their data halfway. The drone industry is currently going through a period of collapse after realizing adoption will take longer to happen than expected; they got stuck dying the long, slow death by prototypes and proofs of concept that we all dread. The satellite imagery ecosystem could easily go through a similar contraction if we keep going the way we’re going. It’s great that you have the government as an anchor client — use that as a foundation to take bigger bets rather than as a weight tying you to the floor.
¹ An earlier version of this post mistakenly said Google Earth launched in 2010 — thanks to Brian Timoney for the correction.
² Thank to Alex Leith for pointing out software is “open-sourced” and data is “openly licensed.” I updated my wording here for clarity
³ I didn’t consult with any lawyers, either. My grandfather was a lawyer, but he focused on criminal law so his advice wouldn’t really have been helpful for this. He also died before the internet was invented, so I’m not sure he understands the nuances of CC BY 4.0 anyway. The original version of this post did not include specific licensing suggestions — thank you to those of you who reached out to asking for specific suggestions and making suggestions of your own. You know who you are, and I’m grateful to you!