Disclaimer: This article is not a professional write-up or a sponsored blog, but rather the ramblings of an Earth observation evangelist who just wants to make satellite data more accessible and easy to use to help us better understand our planet and make efficient data-driven business decisions.

Recap

In my first post, I explored the analogy of the Video Streaming sector and the Earth Observation (EO) sector focusing on the similarities in search costs involved for customers in finding content in streaming providers and finding data in EO data providers. In my second post, I wrote about the significance of evangelizing EO and how it is critical for its long-term growth, more so than for its short-term revenues. In this post, I would like to explore if there can ever be a Google for Earth Observation and how far away are we from this, leveraging on some learnings from my previous experiences in the software industry.

A Google for Earth Observation

In my first job, I was writing code and developing software. In my second job, I was prospecting and selling software solutions. Then, I switched careers into the space industry and in my most recent job, I was a strategy consultant, focused on the Earth observation (EO) market. So, naturally, I have been asking myself this question: How far are we from creating software apps based on EO data? Those that could be used by mainstream consumers and enterprises alike to get answers to different types of questions, depending on their needs. Maybe a queryable interface, where one can go and type a question in and get results powered by EO data among other data? Like a Google for Earth Observation.

How far are we from using a Google Earth-like application to search: “how many trees are there in Paris?” and it then shows you the results based on EO data fused with other sources of data — the number of tress, the type of trees and a graph of the change in the number of trees over the years and so on.

Or in software terminology, when will we unlock the “Application Layer” of Earth observation? The figure below shows a simplified software operating stack with an example of how Google’s software stack works.

A (Very) Simplified Software Operating Stack & The Google Example

Application Layer: The layer where the users interact with the computer to ask a question and get the results. The application layer also serves as the “abstraction level” to hide away the complexities of how Google works (so that we would never have to type a complex machine readable query to get the results and learn the technicalities of how the search works)
Computing Layer: The layer where the processing of the questions happen with the question converted into a machine-readable query (the actual algorithms and the complex commands & queries).
Data Layer: The layer where the collection and storage of data happen, mostly aimed at making preprocessing easy i.e. optimise the data layer to facilitate the processing done at the computing layer to return results quickly back to the Application Layer.

I would like to apply this operating stack to the Earth observation sector. Why? Because I think satellite data is just another source of data (like data from social media, economic data, data from IoT / other sensors etc.), whose sole purpose is to serve as an alternative data to create scalable software applications. Unless satellite data is able to unlock this application layer, my guess is EO will never reach that multi-billion dollar potential that we keep talking about.

How does the Earth Observation Operating Stack look like?

The current state of affairs mean that the EO Operating Stack turns out to be incredibly fragmented — a variety of companies serving different verticals, with very different business models. Does it really have to be?!

Data Layer: Includes the Collection layer who do all the hard and cool stuff of building satellites to collect data and the Aggregation Layer who help organise and disseminate the data that is collected from space.

Collection: Incredibly fragmented with plenty of players trying to acquire the respective market based on the type of the data they collect — the optical data market, the radar data market, the hyperspectral data market, the AIS data market, the infrared data market and so on. Interestingly, most of these players are becoming vertically integrated and hence they (ideally) want to operate in more than just the Data Layer, not just providing Data-as-a-Service. But, is that realistic with so much competition and no collaboration among them?! Also, why are some data providers still using the obsolete FTP to transfer gigabytes of satellite imagery? Where are the mobile apps for monitoring and tasking satellites?

Amazon Prime, delivers within a few hours these days, but we take a few days for acquiring satellite images. Why shouldn’t acquiring satellite imagery become as easy as buying a book on the Kindle and be delivered seamlessly?

Aggregation: Given that we have so many types of data, it is obvious the EO market has / needs an aggregation layer to save people from the trouble of “accessing” data. The solution? A number of providers trying to become the Netflix for EO data powered by the cloud. But here is the core assumption of this layer: A vast majority of users would like to “stream” EO data and will use the platforms for acquiring and developing applications. But, do we really have so many users with remote sensing expertise, who not only know how to process geospatial data, but are also willing to understand what spatial resolution, radiometric resolution, nadir etc. mean, whilst spending money on both imagery and cloud services?

Computing Layer (Pseudo-Application Layer): The two enabling layers, Analytics and Insights, take up the trouble of acquiring & processing satellite data and packaging the analytics and insights for end users. However, this is still not the application layer, although some products here tend to bridge that gap and cross the chasm into proper SaaS solutions.

Analytics: This sub-layer consists of players providing ready-to-integrate analytics leveraging on advancements in AI and computer vision providing results in a vertical agnostic fashion — cars, trees, buildings, aeroplanes, roads, ships, containers etc. So what? As an outsider inside the EO industry, this is what I find myself asking, “so what?” It’s truly amazing that we can detect these objects and count them from space. Sure, there are a number of users who might capitalise on these analytics and build specific tools internally for supporting their business processes. But, larger market of non-traditional end-users and enterprises, are we really them to figure out what to do with these analytics? Maybe I am wrong, but aren’t we supposed to evangelise the technology and spoonfeed the solutions to the niche markets?

Insights: Being the closest thing to the application layer, and sometimes, even acting as the application layer for the EO sector, “insights” are what the end users truly care about, somewhat in response to the “so what” question of the Analytics sub-layer. But given the domain expertise needed for providing specific insights, it is obvious that we will have different players focusing on different verticals. But, is that really the most optimal way to provide insights? Are we only ever going to have vertical-specific software insights? Can there not be a generalist software application based on EO? Or perhaps, a plug-and-play tool powered by EO for widely used enterprise software such as Salesforce, SAP, Oracle etc?

Are we doing enough to abstract away the complexities and technicalities of EO from users who do not want to reinvent the wheel, who do not care that space is sexy and who just want a piece of software that makes their day-to-day job easier?

Application Layer: The elusive application layer, both aimed at the consumer apps and enterprise software. If Google can be the search engine for all types of questions and Salesforce, the CRM for all kinds of verticals, why can’t we have enterprise software powered by EO? Also, do B2C apps have no scope for the future of EO? Are we limited only to sharing beautiful satellite images? I am a big fan of Planet Stories, I thought it could become the YouTube for EO, and is the start of something new, but does not seem to be going anywhere. The closest success story for B2C that I can think of is Daily Overview, which sells print copies of satellite imagery. I am also looking forward to Sen, to see what type of consumer apps they might unlock with video data from space.

What are we missing?

To be able to build mainstream applications, we might be missing something. The first obvious thing that comes to mind is: demand. Where is the demand? Who is going to need a Google for Earth Observation? What is their willingness to pay? Sure, we need to try and start from some use case and some vertical, which is why I am excited about the “Insights” layer, and the Insights-as-a-Service model, which has the perfect leverage to cross the chasm and graduate into a SaaS solution. But, in order to get there, here is what I think we are missing before we can unlock the full potential of the Application Layer.

Gaps in Data: Even with my amateur remote sensing knowledge, I would be lying if I said we have all the data that we need to build applications for any problem and hence, the challenges are only in the computing and application layers. No. Hyperspectral data (Hypersat, Satellogic) is going to be huge for predicting crop yields and detecting new mines. Daily revisit with very high resolution data is a necessary, not sufficient condition for better object detection. Infrared data (ConstellR, SatelliteVu) could potentially be huge for green energy transition through building insulation and heatmaps. Videos from space might become the next big thing for change detection and starting a B2C phenomenon. So, we definitely have gaps in the Data Layer that need to be filled, through existing players or new ones. I foresee that all the companies in the Data Layer might be on equal footing in a few years, and the competition will perhaps not be about what type of data/what resolution they offer, but rather what applications they are unlocking.
Handling EO Data: The world is not ready to handle gigabytes and petabytes of EO data — clearly our in-built GPUs are not and the costs of paying for acquiring, storing and computing potentially useless, cloudy imagery in the cloud. So, in order to handle EO data better, we would need to start processing it in-orbit with edge computing (OrbitsEdge, KP Labs, Spiral Blue). Big Data Overload is a thing that most enterprises are worried about, and hence, we cannot expect our end users to take care of handling all the incredibly useful and useless data we are downlinking from space. They might give up on us sooner than expected.
A Broken Data Business Model: Joe Morrison wrote a brutally objective article on why the satellite imagery is broken, causing quite a ruckus in the industry. It was the elephant in the room and someone finally said it, the business model status quo was not sustainable.

Why aren’t we “commoditizing the complement?”

Microsoft did not become successful because it locked away its license for MS-DOS in a locker or had an exclusive partnership with IBM; it licensed MS-DOS to every single manufacturer with the goal of commoditizing its complement: the computer hardware. Soon enough, the prices for hardware went down, competition in the hardware manufacturing market went up, but guess what every manufacturer was using? MS-DOS! Isn’t this exactly what is happening with EO? Prices of imagery going down, competition going up, and more importantly, substitutes coming up (stratospheric-EO). Which is why it is critical that companies in the Data Layer open up their archive of satellite imagery and let us build scalable software applications (the application layer) so that the complement — EO data — can become commoditized and hit that multi-billion dollar market size figure.

Conclusion

Clearly, things to be fixed are in the Data Layer, both technologically, to fill the data gap & process data on-orbit and commercially, to fix the broken business model. Going back to my example of a Google for Earth Observation, where I can type in “how many trees are there in Paris?” to see the results, we might not see the light of its day immediately. We have too many smaller, specific applications to build, which is carefully being done through the Analytics and Insights layers. But, I do hope that we don’t lose the bigger picture of making EO mainstream, with the ability to create and/or integrate into enterprise apps and attract general consumers. Ever since I got into EO a few years ago, I have been truly amazing at this new source of data from space and become its evangelist, of sorts. So, it’s only fair that this evangelist has a vision of seeing EO become mainstream, like a Google, a Salesforce, an SAP or more realistically, an indispensable underlying powerhouse for these big software applications, used across industries and countries. Am I wrong and completely out of my mind? Or, is this wishful thinking?

Note: Planet had set its Mission 2 to be Queryable Earth, which I guess could probably fall into the same basket as becoming the Google for Earth Observation. Even though, I am unsure if they can become a Google for EO with just optical data, kudos to them for making an attempt!

TerraWatch Space Insights

A Google for Earth Observation? Perspectives…