Commercial datasets done (mostly) right — always fresh

Seamless access to commercial data and what are the preconditions for that

Grega Milcinski
Planet Stories
6 min readJan 25, 2022

--

Arecibo Observatory radio telescope being dismantled, processed by kosmi (credit CNES/Airbus Space — Pleiades in November 2021)

A year and a half ago we have introduced a first set of commercial data providers available via Sentinel Hub — Airbus Pleiades, SPOT and Planet PlanetScope (later added Maxar WorldView and GeoEye as well). The feedback from the community was immense. However, it also became clear how much of a disconnect there is between non-expert satellite data users (vast majority of them actually) and the standard VHR offering. There were some non-realistic expectations — i.e. hoping to get access to full archives within Sentinel Hub standard subscription packages or that they will at least be able to “look” at the data free of charge (at full resolution), before purchasing and downloading them. There were (and still are) some issues that are a consequence of the marketing of the commercial providers. These can be summarised in “Why is there no data?”. Most VHR providers promote “daily revisit time” but don’t point out this is only a capability of their tasked constallation, and even then it is only realistic for high priority orders. If someone comes and tries to get past data for a random point in the world, they will quickly realise there are only a few sporadic observations available over a year, sometimes none at all (Las Vegas does seem to be an exception though, having super-dense coverage — probably a luck due to the fact there is not much else in the area). Therefore, if one wants to monitor the area, they need to order tasking and this immediately becomes super expensive with minimum order sizes of tens or even hundreds of km² of homogenous area. Weekly coverage over a year will quickly exceed 100k EUR. And if one wants some historic coverage, bad luck. With this rant past us (not unlike the Joe Morrison’s one), the rest of the post will cover two topics — first one on integration of PlanetScope data, that comes as close as possible (or better) to the coverage our users are used to, and the second one on what we believe commercial providers should do to make the data more widely used (assuming they want to, obviously).

MSG Sphere Project at the The Venetian Resort Las Vegas, ©CNES 2019–2022, distribution AIRBUS DS, Pleaides data processed with Sentinel Hub. Two years of data, 458 scenes.

Fresh data always on

What is so special with PlanetScope? A couple of things. First one is that the data are systematically acquired (almost) every day for every part of the landmass regardless of whether there is a customer for it or not. “Almost” due to the constraint that if observation is too cloudy, it will not be distributed to the users due to insufficient quality of the ortorectification. Here the “daily revisit time” is not just a possibility, it is the reality. Second, there is Planet’s “Area under management” (or “Hectares under management” as we call it) business model, for which user is paying an annual subscription for a specific geographic area and gets access to both archive and fresh data, without additional charges. This model is not appropriate for all the use-cases — if one is only interested in discrete imagery over a couple of tens of km², it will typically be too expensive. For agriculture monitoring (and similar), however, the model works very well. The last component of what makes Planet standing out is their distribution platform. Where it often takes several hours or even a day for other providers to make the data available, Planet is able to process desired scenes in a couple of minutes.

With all three components being addressed, we were thinking on what to do in order to make the data accessible in a truly seamless fashion. In general, to access commercial data in Sentinel Hub, the process is two-fold. One first has to “order” the specific scene and wait for it to be ingested. Only then can our users stream it in their applications. The reason for this type of implementation is clear — every time the user makes an order, there is additional cost incurred so one has to be careful. But with PlanetScope and their AUM/HUM this is not really the case. Here is where the new shiny thing comes — Subscriptions API. Our users can now mark an area, state “I am interested in the data over this area”, do this action only once, and from that moment on (until the user decides otherwise), the PlanetScope data will automagically be available. Exactly like for other monitoring missions, i.e. Sentinel, Landsat, MODIS, etc.

Check how to use Subscriptions API in our FAQ.

Wishlist for all the newcomers in the commercial space

There are surprisingly many commercial data providers out there. We get introduction letter several times a month from the companies who would like us to resell the data for them. It’s, however, not often that the resale process is successful. Here below we are listing few things that we would hope the providers would address. Not just because of Sentinel Hub. We believe that majority of the users, for sure those called “value adders”, have similar needs. At least those in the “general domain” where many providers would like to break into. Perhaps a useful action list for all the startups who just raised millions of dollars to build a plethora of satellites, but have not yet started to think on how to get data to the users:

  • Transparent pricing — everyone would like to see the prices being as low as possible, as it is easier to find end-user willing to pay for the service. What is even more important, though, is that these prices are predictable, however low/high they are. Nobody can build a business model on the data if they don’t know how much it will cost. Also, requiring for each purchase to exchange several e-mails to get the quote, is somehow bizarre in the modern world. A very important part of the pricing is “minimum area of interest”. There is clearly some overhead involved with each order (for tasking even more so, as one has to move satellites), but very few people need 100 km² large area. This criteria is typically impacting the final price more than the “EUR per km²”.
  • Clear license — the “internal use license” is super limiting, but more than the limit, it is confusion that is problematic. Who is the “end user” in the world of streaming services, value adding applications, etc? It should be the last user, who is actually looking at the data (or making a derived product). It would be great if the “multiple user licenses” and “media licenses” and “non-commercial use licenses” would get standardised somehow, possibly reflected in the pricing. The users, whomever they are, can then come, pick a license based on the cost/constraints, and use it.
  • Stability of the business model — it takes quite some time, for us and our customers, to integrate new imagery source and build a business on top of it. Six months at least, if not a year, for actual revenues to start trickle in at decent pace. The worst thing that can happen is that the business model changes in this period — either significant change of prices, constraints in how to use the data, or similar. We should all think long-term or not at all.
  • Territory rights — country-based exclusivity is becoming a thing of the past. In the world of Internet and web services, we are selling to users across the world and they are providing their apps to end-customers globally as well. If we are somehow limited in to whom we can sell the data, it’s causing significant friction in the business model. There are exceptions, such as sanctioned countries, but these should be an exception rather than the rule.
  • API — if we want to build a scalable business, we have to get rid of just about all manual interactions. Having a reliable machine-to-machine interface on all sides is essential component. E-mail notification that “data is available on FTP for download” is not it. Check how Airbus or Planet have done it. Check our API. Build on top of STAC. We can accommodate to specifics (the more standard it is, easier it will be, though — for us and others). But do have an API, well documented one, with examples and possibly a sandbox solution, which developers can use to perform integration without racking large bills for data they do not really need.
  • Standard formats — COG is the most commonly mentioned these days, but JP2 or others will work as well. Just make sure that all deliveries have the same structure. If you are using COG, you might even use our “Bring your own data” facility and push data directly to the users, without us in the loop.

Disclosure: We have recently agreed on cooperation with Planet, but the thoughts above were compiled before that and did not change after this fact.

If you want to learn more about Sentinel Hub, make sure to listen the MapScaping Podcast:

--

--