Back in the mid-1990s, when I was a young research scientist in machine vision and artificial intelligence at the Norwegian Computing Center, one of my tasks was to analyse satellite images of the Norwegian mountains. The objective was to estimate the amount of snow in the mountains during winter. This was done in order to understand the risk of flooding during the spring. The data also had another interesting use, as there was a correlation between the amount of snow in the winter and the amount of water feeding Norway’s 278 hydropower plants, and therefore also the future production volume and cost of electricity.
The number of satellites orbiting the earth has undergone an enormous rise over the past decades. Over the past few years there has been a multiple order of magnitude drop in the price of satellites and a commensurate increase in the availability of the imagery they produce. In the past those images were just available to governments, but now pricing has come down to make satellite imagery accessible for a number of commercial uses. As pricing drops further, I believe aerial imagery from satellites and drones will be a new and commonplace data source for next-generation business analytics.
One company that has taken satellite imagery to the next level is Palo Alto-based Orbital Insight. Using advanced image-processing, machine vision and cloud-based computational power, they use satellite images to determine a wide range of interesting commercial insights, such as estimating retail sales by counting cars in shopping centre car parks, creating independent data about the health of the Chinese economy by measuring the amount of commercial construction work, predicting the yield of crop harvests by tracking agricultural fields and many others.
According to founder and CEO James Crawford, we already have the ability to take 8 million square kilometres of imagery across the world every day, and that will increase by a factor of ten in the near future because of the number of private satellite start-ups entering the market-place, and by another factor of ten as drones become commonplace in our skies and provide better-quality images than satellites.
The new entrants are building satellites that are incredibly small and cost a fraction of what they used to. The growing array of satellites and drones will mean that eventually we will have access to images of every city in the world at every moment in time – a volume of data that it will be impossible for human beings to process. Consequently, analysis will be done by machines. Deep learning and AI will scale up our ability to look at images and will be able to spot geo-economic trends across the world.
Big box retailers, which have large parking areas, are rich potential sources for accumulating powerful consumer data that can be used for multiple forms of extrapolation. For instance, Orbital Insight can offer its financial services clients predictive data about the quarterly performance of Walmart or other big box retailers by looking at images of car parks. Aggregating several years’ worth of this data can produce a heat map that shows where shoppers prefer to park, and other trends such as seasonal patterns of behaviour and other time frames, like days of the week. It’s possible to do comparisons to judge which of two competitors is performing better – data that’s extremely valuable to investors. According to Crawford, activity in the car park is directly related to that company’s stock price.
Aggregating a large amount of data enables us to see macroeconomic trends that offer us insight into the performance of the wider economy with a high degree of accuracy because of the scale of the information. Orbital Insight aggregates data from fifty retail chains across the US in order to get a macro view of the US economy. Commercial drones will increasingly be used for this purpose as well.
As we move forward, Crawford believes that, as well as using this kind of data for financial forecasting, we will understand and contextualize store performance by understanding general trends, understanding customer behaviour such as whether consumers are having trouble getting to a store, the influence store locations have on sales, traffic patterns within cities and regions, and anticipating supply chain disruptions such as bottlenecks in ports or transport problems with major suppliers. Understanding the world as a geo-spatial problem – whether this is images from drones, mobile phone counts or car counts from connected vehicles – that can be analysed at scale can provide crucial data to industries including retail, energy, insurance, health and finance, not to mention government applications.
One start-up that is pushing down the price of commercial application of satellite imagery is Planet Labs, based in San Francisco. The company, which has just over $151 million in venture funding, is an aerospace business that uses off-the-shelf materials to develop and build low-cost imaging satellites, known as Doves, that are little more than the size of a brick, with a weight of about 9 lb. These satellites are sent into orbit as passengers on other missions, attached to rockets, and are for this reason much more cost-efficient to deploy. Each Dove satellite continuously scans the earth, sending data once it passes over a ground station. Together Dove satellites form a constellation that provides a complete image of the earth at 3–5m optical resolution. The images gathered by the Dove satellites provide up-to-date information relevant to climate monitoring, crop yield prediction, urban planning and disaster response.
Planet Labs has a very different model from that of a government organization like NASA. Although not directly comparable, NASA’s Landsat 8, which was launched in February 2013, cost $855 million to develop and is the size of a truck.1
Since its start in 2010, Planet Labs has designed, engineered and launched seventy satellites into space, more than any other company. Once it has 150 in orbit (expected to happen in 2017), Planet Labs claims that it will be able to send back images twice a day that take in the whole of the earth. This avalanche of imagery will create an unprecedented database of the entire planet, one that can be used to stop forest fires and maybe even wars.
There are a number of other organizations mapping the earth from space, including Terra Bella, which is a subsidiary of Google. Its satellites are around the size of a mini-fridge – like Planet Labs’ satellites, they’re built with off-the-shelf components – and send static images and HD video back to earth, where they can be used to understand the movement of trucks transporting products from, say, a distribution centre to a retailer, the amount of wattage coming from a developing nation where uptake of electricity is spreading or the amount of a discolouring pollutant in the bay near a city.
All of this data has governmental and private applications and is as relevant to scientists and environmental campaigners, say, as it is to economists and analysts at financial institutions building forecasting models. If you can examine oil storage tanks from above, you might have a sense of how much is being pumped and added to the world market. If you can analyse the number of trucks coming out of Foxconn’s manufacturing facility in Shenzhen, you would have an idea of when the next iPhone is going to be released.
In July 2016 the Japanese tech investor SoftBank announced a $32 billion acquisition of the British chip producer ARM.2 The bid was a staggering 43 per cent higher than the last closing price and 41 per cent (!) higher than its all-time high.
The acquisition represented SoftBank’s belief in the future of the Internet of Things (IoT) and was an investment in a future transformational technology trend estimated by a 2016 World Economic Forum report to create a value of $19 trillion over the next decade in cost savings and increased profits.
It is hard to grasp such immense value creation, but whether the report from the World Economic Forum is accurate or not, it is pretty clear that the IoT will impact the world in a very big way.
The Internet of Things can be described very simply as a large volume of interconnected sensors with processing powers. These sensors can be embedded in almost anything, almost anywhere. Imagine a light bulb equipped with a sensor that can detect that the bulb is broken and send this information to a janitor who knows where to find a new bulb and what equipment is needed to replace it. Such sensors can be utilized in manufacturing to create efficiency and automation in factories; they can add much more precise data to logistical processes and create a lot of value for processes and businesses we currently cannot imagine.
The interesting aspect of the IoT with respect to Outside Insight is the new data it collects. Admittedly a lot of the IoT data will be company internal data, where it can improve a lot of operational decisions and processes, but there will also be a range of publicly available IoT data that companies can tap into. This is illustrated by some of the smart city initiatives deployed on an experimental level in cities such as Amsterdam, Barcelona, Stockholm and Singapore. As part of these cities’ ambition to create efficiency and improve their citizens’ quality of life, interconnected smart sensors are deployed widely to identify traffic congestion, optimize power consumption and improve public safety. In the process, a lot of information is collected and aggregated. It is unclear how much of this information will be publicly available, but as sensor technology and processing power become cheaper, one can easily imagine a future where every street, every house, every traffic light and every road junction is littered with sensors collecting data that can be used for analytics.
Satellites and drones are swarming the sky, and on the ground tiny sensors are swarming our homes, our bodies (in the form of wearable tech), our vehicles and our surroundings. When combined, they provide data about imagery, temperature, humidity, pollution level and a whole range of other detailed information.
From an Outside Insight perspective, the Internet of Things will in the future provide a new rich data source that companies will be able to use to predict customer behaviour, future demands, the success of their competitors and a whole range of other insights that today are hard to imagine fully.
The wealth of information available on the internet today is mind-boggling. And with every day that passes it continues to grow exponentially. And that is before the Internet of Things really takes off. As new sensor technologies become more prolific, virtually anything will be measured and logged. The Internet of Things alone will probably produce as much information as all the information published on the internet. And as drones and satellite imaging develop further, virtually every spot on the globe will be monitored and recorded in video, sound, and infra-red.
We have a lot of data today, but it will be dwarfed by the data we will have going forward. The data will continue to grow exponentially. With the growth of the data, more insights can be retrieved. They will increase the potential value of Outside Insight, as long as we develop technologies that enable us to analyse the vast data sets we will be dealing with.