Shifting Gears
Punch the clutch to change gear, but raise it gently to find the biting point. EO-derived analytics are still finding the bite.
I was lucky enough to learn to drive in the UK, so I learned to drive with a manual gear-box. On a stick-shift, for those in North America. Now, living in Canada for almost two decades, those clutching skills atrophied until last year when I rented a vehicle for a family holiday in Greece. I rapidly recalled my hill starts in a sweaty moment on a steep rural switchback with a honking truck bearing down on us. No stall; we survived.
Hill Start
There is a point, though, when switching gears, when the clutch is depressed and the engine is disconnected. That's obviously the point of the clutch, so the engine and axles aren't torn apart by the change in gearing. I think that's where we are in Earth observation, indeed in geospatial at large: a gear change. The engine is disconnected, and our industry is feeling the biting point.
To me, this pause is due to a couple of industrial patterns. One is modernization, and another is about new innovations finding their place. In this article, I want to consider what modernization needs to look like for robust enterprise companies.
I have spoken frequently about innovation curves and the transition between them. Witnessing the shift between curves in real time is interesting.
Joking with many about the notions of modernization and digital transformation, with my tongue in my cheek, I would scoff at the need for companies to transform for changing needs. But in reality, this transformation is happening around us in real time, and much more is coming. It’s not a fair joke.
We can see how different companies view these innovation curves and want to position themselves accordingly. Most companies may want to appear "innovative," but in reality, they won't move their core value creation processes until they absolutely need to, when they have complete faith in the move. While it’s important for an organization to sense, it’s more important for that organization to persist, and that means not interrupting any of that value creation.
This is an obvious statement feeling a bit like "if it’s not broken, don't fix it.” It is also completely reasonable. An organization that has created value in a particular way will be reluctant to switch up that process on the whim of a startup or consultant. So, when Earth observation people start providing whizzy new analytics and pixel-based products to established industries, big surprise, there is some initial curiosity, but then the clutch is depressed, and the engine disconnects while the enterprise determines if they really need to care.
Getting off PoC island
The big gap for geospatial technology has been between proofs of concept and practical enterprise implementation. This is a surprisingly wide gap, even in the US Government (USG) environment. We could argue that both Descartes Labs and Orbital Insight ultimately fell into this gap. I commend both organizations for attempting the leap on behalf of the geospatial community, but they seem to have fallen afoul of the industrial pause in commitment to moving from PoC to operational workflow.
What causes this pause? If I am to make two key observations, it is that:
The industry has to be ready for the change.
The change has to be impact-free.
Readiness is measured against inertia. An example of what I mean is that many organizations didn't want to commit to anything cloud-related until COVID forced that change. Then, it happened fast. The tectonic societal shift provided an energy greater than that of organizational inertia. It forced movement. If left to wait, the organization must grow to become comfortable with the change.
Building that necessary energy can take a long time without a forcing factor. Often, these longer shifts are forced by major industrial changes or licensing, such as Adobe’s switch to SAAS.
The other observation was impact. We often talk about having a "Big Impact," but, in fact, for enterprise companies, innovators want to have almost no impact. What I mean by that is usually, large organizations do not want to change workflows, but they do want their present workflows to be better. There is a delicate nuance in that difference.
Getting a large enterprise to change a workflow to access your exciting new geospatial SAAS will be near impossible. First, IT will insta-block you with authentication and security concerns. Then, middle management will block you because they don't want to change a functional workflow (don’t underestimate the reality of career risk: Would you trust you?). Then, you'll have to retrain a team with a new toolkit, all against the pressures of making quarterly results.
Instead, innovation needs to be almost invisible but just much better.
Think about files; not apps.
Think about workflows; not toolbars.
This sounds kinda old skool and anti-innovation, but it’s not. Expecting a Fortune 500 to log in to your independent yet innovative geospatial SAAS is crazy. But adding value to a workflow by making a process faster or easier makes tons of sense. Do that, and do it invisibly. You can have a huge impact, by minimizing your innovation’s workflow impact. This of this as a ratio of enterprise productivity to workflow impact. Ensure your productivity benefit is a big multiplier and the workflow impact is a tiny number.
I hope that these two pointers help with your industrial modernization projects. My team at Sparkgeo has lived these experiences, and we continue to do so every day. We’ve come to realize that it helps to meet people where they are.; organizational inertia is a real force in business. You can accept it and harness it, or you can be continually frustrated by it.
Industry downshift or up?
It also feels as if our broader geospatial community is in the midst of a gear change. The space SPACs have impacted the spirit of the smart space EO sector. Clearly, the stock market at large expected much more revenue than has materialized, and the messaging of "just wait, it’s coming." doesn't seem to be working anymore. To be fair, it never worked. Now, venture-backed EO and analytics companies are having trouble raising finance rounds because the exits aren’t so obvious anymore (of course, it’s about the exits!)
As alluded to earlier, the two biggest venture-funded geospatial analytics platforms of the last half-decade have met with mixed acquisition ends. Each found itself in interesting circumstances, but neither ultimately grew into the behemoth we all hoped for (yet). The GIS sector has largely stagnated around Esri technology, and as with any industrial monoculture, innovation and hybrid vigour are impacted without competition.
Hit the gas
That’s all doom and gloom. And, to be honest, the last twelve months have felt troubled. Private Equity has picked up major geospatial institutions (not a failure, but PE looks for opportunity). And consolidation has been in the air. But I am nothing if not an optimist (I think most entrepreneurs need to be, but it's a character flaw).
Last week, I attended both GeoIgnite and SatSummit. During these events, I was lucky enough to meet amazing entrepreneurs, scientists and policy people. In these discussions, I was struck by the need for EO data to be consolidated meaningfully. That consolidation needs to make EO useful beyond itself; EO data is still hard to use. As I have discussed numerous times, pixels are each a problem to be solved, so we need to get beyond those pixels. I will discuss this subject more in the future, but much to the surprise of my team1, I think Large Earth Foundation Models, like Clay, hold the key.
The idea of going from "Pixels" to "Analytics" has been around for a long time. To be honest, the use of the word “insights” has become a bit of a trigger for me. Unfortunately, every analytic quickly becomes very specific: "I have a great model for counting domestic swimming pools in Huntsville," but it won't work in Seattle. So, tuning a model or algorithm for every geography becomes a game of whack-a-mole. The foundation model approach should solve this. Indeed, there is even a possibility that the foundation model approach addresses some of the components of the knotty Analysis Ready Data problem. Which I think of as Earth observation’s dirty little secret.
If we can solve this, then we get to press that gas pedal. We get to harness all that data flow and toast the end of the Lament for the lonely pixel.
My team knows that I kinda hate generative AI. I think it creates lazy and uncreative people. I hate the number of meaningless and wrong listicles written by ChatGPT for clicks, which subsequently get pulled back into the model to reinforce wrongness. Obviously, I know those articles are posted by people, and it's not actually ChatGPT’s fault. My beef is mainly with those lazy, click-seeking individuals. However, I also concede that, like most things, generative AI and Large <fill-in-the-blank> Models can be incredibly powerful in the right circumstances. This may be one. Indexing corporate and alternative data sources may be another, and there will be numerous more. I am generally pro-AI, while being disappointed by people.
Brilliant article Will, your sense of sense of reality is a breath of fresh air in our industry