Cover Story: Data mining from space

This article first appeared in Personal Wealth, The Edge Malaysia Weekly, on October 2, 2017 - October 08, 2017.

Satellite photography is used to monitor the growth of crops

-A +A

The use of “alternative” data is gaining momentum as a means of getting ahead of the curve and making better investing decisions. In particular, satellite imagery — coupled with intelligent, cloud-based processing techniques — is extracting extra value, especially for fund managers when they assess potential investee companies.

Geospatial analytics is the evolving intersection of big business and big data, which allows fund managers, family offices, investment holding companies, insurers and even private citizens to tap unprecedented amounts and types of data acquired from the hundreds of satellites in orbit around the planet.

This class of data is playing an increasingly important role — as a complementary data set to be read alongside reports by investment banks as well as an evidence-based document that can provide clarity to, or correct, analyst reports.

For example, as Hurricanes Harvey and Irma tore through the US, hundreds of satellites snapped “near real-time” images of the storms and the devastation they left. In the immediate aftermath, some insurers purchased access to the high-quality satellite imagery, replete with added analysis and intelligence. The information was used to ascertain damage, as well as calculate costs and claims for their customers.

On a small island off the coast of Singapore, eyes in the sky monitor crude oil levels in vast tank farms. These huge oil storage tanks utilise floating roofs that rise and fall with the levels of crude in storage. The data was purchased by funds looking to make risk-adjusted decisions on whether to invest, increase or divest their shareholding in that energy company.

By tracking outdoor car parks at a large multinational retailer, for example, fund managers are better able to predict share price movements before the retailer produces its quarterly report. As it turns out, monitoring traffic movement in car parks is an accurate proxy for footfall.


Welcome to the world of satellite imagery

Geospatial technology combines high-quality satellite images with machine learning engines and big data analytics software. These automated processes churn out actionable insights that can be customised according to a client’s industry and other specific requirements.

One company making gains in this new sector of data analytics is Geospatial Insight. The British outfit, co-founded by Dave Fox, recently completed a Series A funding round, with the equity investments led by venture capital firm VenturesOne. Fox, who is the CEO, speaks exclusively to Personal Wealth.

“We have had direct feedback from one of our clients — the trading arm of an oil major — telling us that the data is making a difference in their trades,” he says.

Fox believes that the unique and evidence-based nature of satellite imagery is catching on in the finance sector. “I have been in the geospatial industry for nearly 20 years, and there is potential for this space to grow. I saw that the market was there and the timing was right, although we were probably a bit early. But over the last couple of years, market awareness has definitely grown,” he says.

The technology is gaining ground in North America, with a report last year in The New York Times that traders at money management giant BlackRock analysed satellite imagery in order to ascertain the health of China’s real estate sector.

CNBC has reported that Orbital Insight — a company similar to Fox’s — worked with JPMorgan in order to track the number of cars parked at American retailers.

Indeed, when retailer JCPenney reported better-than-expected second-quarter results in 2015, the news surprised all but a small group of investors. This group of mostly hedge funds had obtained satellite-derived intelligence on JCPenney’s car parks across the US.

These select investors were able to tell in near real-time that traffic into JCPenney stores rose in April and May, getting a decisive jump on the market.

Players in the geospatial sector are also newly awash with cash. Last month, US player Descartes Labs announced a US$30 million Series B funding round led by March Capital. Only three months earlier, Orbital Insight closed a US$50 million Series C funding round with Sequoia Capital.


A bird’s eye view

As more satellites were launched over the last five to six years, satellite imagery eventually became a viable tool to monitor daily events and activities on the ground. This was a key turning point for the technology, one which is now driving a gradual proliferation among finance professionals.

With the greater number of satellites in orbit came economies of scale, resulting in big drops in cost, making the technology more accessible. Fox says: “Now that the pricing model has changed, we can produce competitively priced satellite intelligence and analyses.

“As more satellites came online, and in particular as cloud technology evolved, we could access archived satellite data via the cloud. This meant that we could survey much smaller areas and no longer had to pay a big premium for a lot of data that we did not need.

“In fact, with years of satellite data now archived on the cloud, we can tap it to see how a situation or event has evolved from the past.”

Geospatial Insight partners a satellite company known as DigitalGlobe, leveraging its fleet of very high resolution satellites. Geospatial Insight analyses the data, adds value to it and repackages it for its clientele.

Fox says a lot of the satellites were initially developed for environmental monitoring, meaning that they are particularly useful at monitoring things that grow. “Now, there is a different class of satellites with multi-spectral capabilities. These satellites can see the world not just as we do — in red, green and blue — but in different infrared wavelengths.

“The human eye would see vegetation in green, because it reflects strongly on the green wavelength but, actually, vegetation reflects much more strongly on the infrared spectrum. So, satellites are able to pick up very subtle things that happen in vegetation, particularly crops.

“We can use this kind of information to monitor how well a crop is developing over a season, or even years. You can tell if it is under stress from disease or from a lack of water. And you can fuse this information with other crop growth data and weather models in order to predict what the crop yields might be.”

With the technology, Fox says, it is now possible to look at growth trends, how the crop has developed and where the crops are in their growth cycle and then look forward to how it might perform and get an estimate of the potential yield.

“To date, we have focused on automating the analysis process as much as we can — we have developed automated tools to analyse oil storage tanks, for example — and we have integrated about 30 different data sets to help predict the yield of wheat crops,” he says.

The company has done a lot of work on palm oil as a commodity, he says. “Palm oil exports from Malaysia and Indonesia make up the lion’s share of the world’s supply, so understanding exactly what is happening in these countries is critical for our clients. We have developed a way to measure the number of oil palm trees as well as how old they are, since the age of the trees is crucial to their productivity.”

He says fields in Johor and Pahang have already been fully mapped, with the rest of Malaysia and Indonesia soon to follow. Geospatial Insight is collaborating with Malaysian and British universities to pioneer a methodology for the prediction of palm oil yield.


Risk factors

As the use of alternative data by hedge funds — from scouring Twitter feeds to accessing geolocation data gleaned from mobile phones and using satellite imagery — continues to gain momentum, privacy campaigners have raised concerns about these emerging data sets.

A Privacy International policy paper noted that hedge funds using location data to inform their trading strategies “had unprecedented population-level insight about individuals’ lives, communities and entire nations and markets”.

Even as more firms wake up to the potential of applying alternative data sets in order to make trades, Fox remains unequivocal about the privacy concerns that might arise from using satellite-derived intelligence.

“When we first started talking to hedge funds and investment banks, this was a key question from their compliance officers. There is an open sky policy in place — one created by a mixture of international law, accepted practice and case law outcomes — which means there are a lot of different satellites up there, and they are all able to gather data anywhere on earth. We can order that data and use it, as can anybody else,” he says.

Privacy issues only arise when, for example, an image can identify an individual or a vehicle’s number plates. “With a satellite however, you do not get anywhere near that level of detail,” he says.

Geospatial Insight does supplement some of its satellite insights with drone images, however, and this brings up genuine privacy concerns. According to Fox, his company edits out people’s faces and vehicle number plates.

He adds, “We don’t always need permission (to take drone images) but it generally helps to do so. If we do need consent, then we make sure to get it before imaging a property. If you’re imaging a broader area at a lower resolution, then you generally don’t require permission. However, there are many other laws to ensure that you operate drones safely, within permitted airspace and with qualified people, which we always make sure we do.

“The rules on commercial drone operations vary enormously from country to country. We don’t operate drones ourselves, but work with professionals who do.”

This is not the only risk factor, however. Fox concedes that even with the vast range of satellites in orbit, the humble cloud formation can still cast a shadow, as it were. “What we do then is use another class of satellite — radar satellites — that can see through the clouds. It does help fill those gaps.”

In addition, the technology would not be suitable to measure the performance of sectors that cannot be seen from the open sky, such as a retail bank, since it is not something that can be adequately measured by satellite images. “It helps a lot if what we are measuring is directly visible,” says Fox, “Anything that grows is good. Also, things like mining operations and oil storage are great examples, and so is looking at car parks, since it has proven to be a very good proxy for footfall.”

Given just how new the technology is for public usage, there appears to be a degree of uncertainty as to whether and how the use of this kind of data might constitute some form of illegality, or perhaps even insider trading.

Fox dismisses claims that satellite imagery might not constitute “publicly available data” for the purposes of insider trading risks. “The fact is, satellite imagery is publicly available. Anyone can purchase satellite data and, in fact, a lot of satellite data is freely available. In theory, anyone can access or purchase the data and create the analysis,” he says.

Ultimately, however, even Fox’s ambitious aspirations are limited by market sentiment. “The main issue is still market risk. This is, after all, a very new technology to a lot of our customers. It is a big issue for us, getting [finance professionals] to understand the benefits and then getting the most out of the data. We find that we are educating almost as much as we are developing business,” he says.


More than just pixels

Gleaning early market insights from satellite images is an exciting innovation, but Geospatial Insight co-founder and CEO Dave Fox says the information is fused with other metrics for added value and intelligence for the client.

Third-party data and metrics form an integral part of Geospatial Insight’s client offerings.

“For soft commodities, such as wheat, we rely on a lot of weather sources. For hard commodities, we provide signals to help inform [clients] on the direction of commodity prices. This includes commodity price data, volume data and how many people are trading these commodities. We purchase these metrics and integrate them into our existing data,” Fox says.

Turnaround times on these satellite-driven analyses are crucial, and Fox says that, ultimately, it depends on the client. Some have far greater time constraints than others, he says. “For example, we have been monitoring the hurricanes in the southern US. In one of these parts of the world, we might have started with nothing, but our service agreements require us to deliver the analysis within 72 hours. So we are literally getting the satellite data the next day, processing and analysing it, and delivering it to the client within 72 hours.

“We have been very busy over the last few weeks, doing work for our insurance clients. We have been using satellite data and even flying aircraft over Houston, Texas, Key West and parts of southwestern Florida in order to help our insurance clients better understand how big the total losses were for catastrophic events like Harvey and Irma.

“But more than this, Insurers are able to use this data to better inform their customers about the damage to their properties, in order to help them settle their claims more quickly. Obviously, for an insurer, it is very important to manage claims efficiently and effectively, and this sort of data can be used to visualise the damage to a given property.

“For the energy sector, we task the satellite the day before. It acquires the data in the morning and it is delivered to us in the afternoon, at which point we spend the day analysing it. We publish the results that same evening or the next morning.

“Our wheat product is a monthly deliverable, simply because what we are monitoring does not change very quickly. It really does depend on the client, and the turnarounds can vary quite a lot,” he says.

The shelf life of the data depends on the client’s request, he says. Trading data would be very important for just a short time. It would be useful for a trader to make an informed trade before the competition gets access to similar data from another source.

“But sometimes, the data has a very long shelf life — historic supply trends and how they have influenced prices over the years could also be relevant,” Fox says.

Now that Geospatial Insight has secured funding, he says the immediate goal is to expand the team. “We grew the team from 8 to 19 in six weeks. The team we have assembled have worked in this sector for many years.”

He adds that a lot of the funding will be used to improve the technology as well as expand product lines. The next phase of the company’s technology drive — over the next 18 months — will be to ramp up automation by way of machine learning, in addition to improving on speed and reliability.

Fox believes the technology could become so advanced that fund manager clients would one day be able to get onto an online platform, click on the specific region they want analysed, specify their preferred metrics and have the information delivered to their inbox the next day.

“However, we would like to think there is still place for analysis on top of that. I think it is still going to need some interpretation for awhile, but the first stage is to make the entire system as robust and rapid as possible,” he says.

“It is now all about getting the analysis in more quickly and packaging that analysis in a way that the client can consume it. Getting it straight into a platform will be difficult, but that is the way this technology will evolve.”