At a Glance
- TellusLabs accurately forecast the 2016 U.S. soybean yield two months early
- Firm provides ag insights at a regional level every day
The problem with data has changed. No longer must institutions consider only how to process large amounts of it. They must think about which data is possible to collect in the first place. Alternative data sources are providing more options than ever. One key resource for financial and commodity markets is the advancement of satellite imagery. When combined with machine learning, it becomes a highly valuable source of intelligence for industries that operate outdoors.
“Agriculture is an example of a sector where the factory floor is playing out under the open sky and yet the market is often moving on impressions and rumors. That’s going to change,” says David Potere, co-founder of TellusLabs, a company that uses satellite technology to gather intelligence and provide forecasts of economically important environmental events. Last year, TellusLabs used its technology to accurately predict the final soybean yield for the United States two months before the USDA released the number. That kind of accuracy paved the way to a $3.1 million funding round in January 2017. Since then, they’ve taken on high profile clients, and taken their product on the road. Potere and some of his colleagues, based in Boston, traveled through the Midwest this year on a crop tour to get the farmer-level view of the fields and counties they observe through satellites. Most recently, TellusLabs was awarded the Futures Industry Association’s Innovator of the Year Award.
Tellus Labs satellite image of a palm oil production area just east of Pekanbaru on the island of Sumatra, Indonesia.
We caught up with Potere at CME Group’s annual Tech Talk event, where he presented on alternative data, its challenges and the possibilities it holds for financial and commodity markets. This is an edited version of our conversation.
Explain how TellusLabs uses satellite data.
At TellusLabs we’re tackling time-sensitive, high-stakes sources of uncertainty that have a market moving impact on the world and are driven by environmental forces. The focus point for us is problems that overlap with earth-scale environmental data, where the only real way you could reduce the uncertainty in a practical way is to use planet-scale environmental data. We’re looking at problems where if you don’t have that kind of data, you may as well not try to constrain the uncertainty.
Agriculture seems like a great place to start then.
We’re really focused on agriculture. There’s way more uncertainty on the global harvest than you would expect given that its $7 trillion of the global economy. The state of play for intelligence information against the harvest, compared to other asset classes, feels like its 30-40 years behind. Our sense is that there is this big modernization coming. Thanks to a combination of new satellite data, geo-located weather records, and more sophisticated ground information – we’re in the middle of it changing the art of what’s possible in terms of getting a clear earlier view of what’s going on in the global harvest.
Satellite image taken of the Rock River in Illinois in October 2016. Infrared reflectance reveals vegetation in green.
Data and surveys on the national harvest through USDA are still heavily used by ag producers. Where does TellusLabs fit in to the data considerations of growers?
At a local scale, we see a great opportunity in low latency. The time it takes for a photon hitting a satellite in orbit to have an influence on our model is quite short, within 24-48 hours. We’re working to get within six hours.
We talked to a corn farmer who is following 12 fields who is saying his goal is to visit his field once a week. That’s the most often he can visit because they’re so scattered. We do see some appetite for low latency coverage at the scale of a whole field. Because we’re standing on more than 17 years of historical data for all fields, we can say things about whether a field is ‘normal’ with a high degree of certainty. That’s a frontier we’re really excited about for grower services, but our focus right now is at county-level and up.
David Potere, far left, with Tellus Labs team members on the Farm Journal Midwest crop tour in 2017.
You’re focused on agriculture. How do you see financial firms using your technology?
We’re not an ag tech company. We’re a satellite imagery analytics company that’s 100 percent focused on agriculture right now. We have a flagship product called Kernel, which is our ag insight platform.
Whether a customer is a hedge fund or asset manager a supply chain officer at food company or a CTA – anyone who has meaningful exposure to the futures markets for corn and soy – they buy a SaaS offering that gives them access to an API and a web app.
Our customers receive everything we have to say about a given harvest — a perspective on crop-masked weather, crop health indexes, seasonal timing, and an outlook on the yield. Instead of forecasting the harvest once a month like the USDA, we do it multiple times a day. So our customers get high cadence, forward-looking models together with all of the other agricultural intelligence in Kernel.
Image near New Orleans, Louisiana in October 2016.
Where is financial services in adopting alternative data overall?
For the larger institutions you are seeing enough recognition on the importance of data that they are starting to centralize and build out their own analytics groups. That’s a big milestone. When companies hit that milestone, they tend to be sophisticated consumers of non-traditional data.
I’d characterize this as early days, especially with exotic environmental data. It’s very different than structured market data. It’s not just econometrics anymore. You really have to understand the planet.
What are the future applications for TellusLabs, beyond agriculture?
We still have an exciting roadmap within agriculture. The team is hard at work extending our U.S. models to the rest of the global balance sheet, starting with South America in a few months’ time. But it’s true that we see opportunity for our technology and approach outside of agriculture.
Timber and forestry management are sectors we think about a lot. It’s just a matter of time before carbon becomes an asset class that has to get audited. Someone has to do the work of assessing how much carbon is locked up in these forests and that it stays locked up. Water is another example of a candidate sector. It’s pretty amazing to see in places like Australia and California that there are increasingly sophisticated market places for water. As we saw in grains and oilseeds, for water we are remarkably under-constrained when it comes to basic intelligence; a topic that is also closely connected to agriculture.