When global food and consumer goods giant Unilever wanted to improve the sustainability of the supply chain in its palm oil production, it turned to satellite imaging company Orbital Insight.
With more than 70 million tons of palm oil produced yearly, the Roundtable on Sustainable Palm Oil calls on stakeholders to move to 100% certified sustainable palm oil, but to do that requires a much more transparent supply chain.
Unilever committed to zero deforestation from its palm oil assets but understood that this was only achievable with the collaboration of all stakeholders at various tiers through its supply chain.
This is where Orbital Insight came in. The company was formed in 2014 with a vision of applying artificial intelligence to geospatial data, working to the vision of founder, chairman, and chief technology officer James Crawford, a veteran of Google and NASA.
Orbital Insight combined satellite images with IoT data sources to build a detailed, global map of palm production which delivered granular visibility to the Unilever supply chain.
This included examining truck parking lots at known locations through the supply chains, such as warehouses, refineries, and ports. This enabled the creation of an anonymized device data which tracked the paths of the trucks, the frequency of their deliveries, and insight into what was being supplied from individual farms.
Meanwhile, analysis of satellite imagery identified any deforestation, and all this data was incorporated into Unilever’s systems to identify the sourcing relationships that would ensure the company met its sustainability objectives.
Connected vehicle data a “game changer”
Mike Kim, the head of Asia Pacific for Orbital Insight, says the company has added significantly to its data sources since it was founded. This has driven more sophisticated and effective use cases.
“When we started, we were only doing satellite imagery,” Kim says. “And then we added mobile location data, and then shipping data, and the latest big one which is coming is connected vehicle data, and this will also be very powerful, and I think a game changer.”
“Within these categories, we have many different data providers, and they change; we swap them out year after year when we find a better data source or someone is lagging a little,” he adds.
“The latest big one which is coming is connected vehicle data, and this will also be very powerful, and I think a game changer”
Also developing rapidly is the AI piece, which applies machine learning, data science, and algorithms to the data. And the third piece is the software platform called Go, which delivers the data and the insights to clients.
“When we started, it was with a thesis to test the hypothesis that companies would pay for this type of analysis because it contributed value to their businesses,” says Kim.
“So we reached out to as many companies as we could to brainstorm ideas, and I have a sheet from the early days of the company when we came up with hundreds of ideas, like counting elephants, for example.
“But over time, we were able to prove that there is an industry here where companies will pay for these types of insights, and our business has certainly developed.”
Another client is the Australian retail consultancy Location IQ, which uses Orbital Insight’s geolocation data to understand the viability of retail projects and make forecasts for potential traffic based on analyzing visitation data from over 1,000 Australian shopping malls.
Retail is one of the industries which Kim describes as a “sweet spot” for the company, along with the financial sector and government with defense and intelligence applications.
Many clients also want to bring their data, combine it with the aggregated data from Orbital Insight, and then build algorithms for analysis.
One example of this is a project where Orbital Insight is monitoring around 100 airfields throughout the world, classifying the types of military and civilian aircraft. There are similar projects underway monitoring key world ports.
“When a customer comes to us and says, ‘can you do this? can you create an algorithm for this?’ we look at the imagery and the data to see if it’s possible,’ says Kim.
“If the answer is yes, then we are confident we can train an algorithm, so the next step is feeding the algorithms with examples to be able to recognize objects.”
Lachlan Colquhoun is the Australia and New Zealand correspondent for CDOTrends and the NextGenConnectivity editor. He remains fascinated with how businesses reinvent themselves through digital technology to solve existing issues and change their entire business models. You can reach him at [email protected].
Image credit: iStockphoto/EvgeniyShkolenko