A Peek Inside CargoMetrics’ Bid To Build The ‘NSA of Global Trade’
Greg Miller, senior editor | 14 March 2018
An aura of mystery has long surrounded CargoMetrics Technologies, a Boston-based quantitative hedge fund that uses ship-movement data to place market bets and has secured financial backing from private-equity giant Blackstone, famed investor Paul Tudor Jones, Google founder Eric Schmidt, Israeli shipping magnate Idan Ofer, and most recently, Maersk Tankers.
The mystery remains, but the shape of CargoMetrics’ business plans and vision is becoming clearer. The company’s founder and CEO, Scott Borgerson, commented on his company’s business model in a presentation at the Connecticut Maritime Association (CMA) conference on 13 March and in comments to Fairplay. Additional information is available from the company’s public patent filings.
“We are not a shipping company. We’re a technology company,” Borgerson told conference attendees. “We’re mapping global trade in real time. Today, my company will process hundreds of millions of AIS [automatic identification system] positions from dozens of satellites and thousands of land-based antennas. We have also geo-tagged hundreds of thousands of [terminal] berths ourselves – which was a lot of work. We have built a dynamic registry of ships. And we’re past shipping: on the wet side, for example, we integrate data on pipeline flows, refinery runs, and the weather.
“What I’m trying to build is the ‘NSA of global trade’,” he said, referring to America’s information intelligence agency, the National Security Agency. “We have a real-time digital map of the global economy. But more importantly, we have a data archive. More than the patent and the first-mover advantage and the dozens of rocket scientists [on staff], this is the biggest barrier to entry [for a competitor]. We save all the data we receive. We have hundreds of billions of historical records in a searchable database. Think of it as a Google search for trade – except it’s secret.”
CargoMetrics is monetising its vast dataset via three business strategies: a hedge fund, a commercial management optimisation business, and eventually, a sale of access to the database to subscribing customers.
Regarding the hedge fund, he explained, “Traditionally, a human trader will try to monetise an information advantage through a sort of gut bet at the casino. We know where every cargo is in the world and we take a purely quantitative approach. We have built a computer algorithm that looks at the cargo flows and statistically forecasts price. It is statistics applied to the shipping field. We manage money, mostly for institutions, and we trade in dozens of paper, liquid markets around the world in real time in a hedge fund that is 100% computer-driven.
“Most hedge funds in the quant space derive alpha and make money by studying price information – it’s called tick data. We think what’s more interesting is fundamental data – having information about the way markets actually are, measuring the world as it actually is, not what the price tells you it is. What the demand really is and the supply really is, measured with actual sensors and satellites and AIS and other means, not what we estimate it to be based on price information.”
To compensate for unknowable variables such as geopolitics, CargoMetrics “hedges through diversification, so we look at as many uncorrelated strands of alpha [investment outperformance versus a benchmark] as we can and combine them in a way that we’re more right than wrong,” he said, explaining that a fund only needs to be right a few more percentage points than it is wrong to reap significant returns. “We’ve been wrong this year 46% of the time. I used to think that sounded bad, but in quant investing, being right 54% of the time is actually extraordinary. When we started [the hedge fund] we had no diversification and we were right 80% of the time until we were wrong 80% of the time – and that caused a drawdown, which was pretty depressing. So we appreciate that we need as many sources of uncorrelated alpha as possible.”
The second prong of the company’s plan involves commercial optimisation, in which algorithms are used to obtain the best returns for a fleet. “This is 2018, an age of the internet and satellites and computers, and most of the world’s wet and dry bulk ships are voice-brokered. You have USD300 billion of steel being directed by humans on phones. It’s no wonder that freight rates remain volatile,” said Borgerson. His system calls for “algorithms to forecast pricing and computers to start making decisions about how to effectively solve a really interesting optimisation challenge. Hopefully, this will be the beginning of a change in an inefficient system in which you have a lot of tonnage being underutilised, and an industry that is fragmented and volatile – because it has yet to be disrupted. We hope to be one of the disruptors.”
This is where Maersk Tankers enters the picture. In July 2017, CargoMetrics and Maersk Tankers entered into a strategic partnership through which Maersk has access to the technology group’s models and algorithms and CargoMetrics helps the product-tanker company enhance its digital capabilities.
Speaking with Fairplay, the CargoMetrics CEO explained, “Maersk is amazing. They’ve made the leap to digital and data, and they really believe in technology. I view them as an operator, no different than how Four Seasons runs hotels and somebody else actually owns the structure. And when you look at returns for tanker operators, they’re all more or less the same. When they flip the coin, they all statistically get the same outcome – because none of them have a true differentiated edge.
“Maersk Tankers is a tanker company like any other, in a very fragmented industry, and they still have humans on the phones running their chartering optimisations and making the decisions, not algorithms,” he said, arguing that “you don’t need a human on a phone to optimise a piece of steel. Mathematics will let you do that.
“We want to help Maersk make a step change in the returns they derive from that steel by taking a dramatically different approach with technology to how they run their business,” he told Fairplay. “I see the volatility of freight rates as an opportunity. You can’t make money in markets that don’t have volatility because they’re perfectly efficient. Freight is not perfectly efficient. It’s a black box that quants have not yet solved – and we’re solving it.”
The third pillar of CargoMetrics’ business model involves the eventual sale of its data. “When you’re building the NSA of global trade, you have a lot of alpha you can trade on paper and a lot of alpha that can be driven by optimising steel, but the vast majority of the data still sits behind our firewall,” he told CMA attendees. “Where we ultimately hope to go is to take a real-time picture of trade flows and put it on the street” as a subscription product.
“We plan to sell it through a couple of partners, although we’re a few years away,” he said. According to the company’s patent filing, “In one embodiment [of the invention], subscribers may access these data through a web-based user interface and/or via an existing distribution network such as Reuters, Bloomberg, or PIRA Energy group.” The patent filing explained that subscribers would be able to set parameters and filters to organise and search data over user defined-time periods; generate value-added outputs, such as average fleet speeds, how weather affects ship movements, the physical location of all vessels in spot trades, port congestion, and early notices of supply shocks and ship diversions; and view the volume and location of an individual commodity across the globe.
“Our ultimate vision it to create utter, total transparency of the supply chain from end to end – a mosaic, a tapestry of what’s happening on the planet in real time,” he said. “To us, this radical transparency is the future. It will be disruptive, but only to people who have been extracting rent from being in the dark corners and keeping the information opaque and for themselves.”
How the CargoMetrics system works
In conversations with Fairplay, several CMA conference voiced scepticism towards Borgerson’s claim that CargoMetrics “knows where every cargo is in the world”. How could it?
According to its patent filing, Cargometics combines vessel data and ship movement data from sources including IHS Markit, Lloyd’s Maritime Intelligence Unit, Clarksons, and Heidenreich Innovations; AIS data from AIS Live, Orbcomm and Com Dev; import and export data from governments and private sources such as Bloomberg and IHS Markit; ship fixture data from brokers; cargo data including bills of lading and cargo manifests; weather data; port data, including information on individual berths, draught restrictions, which berths handle which cargoes, and cargo loading/unloading times; and other information.
The challenge in using AIS data to determine cargo movements is that many ships carry many different kinds of cargo. Fairplay asked Borgerson how his company could know which refined product was aboard a product carrier, or which bulk commodity was loaded aboard a particular Handysize bulker. “I’m not going to give a newspaper our trade secrets, but this are solvable problems,” he responded, pointing to the public patent disclosures and explaining that the answer is related to statistics.
The patent documents imply that at least to some extent, CargoMetrics deals in probabilities, not exact knowledge. It said that CargoMetrics’ system uses “rules-based logic, Bayesian logic, neural networks, learning algorithms, or other mathematical methods” to interpret and integrate all the data.
“Certain ships carry multiple cargoes,” said the patent filing. “This invention resolves that issue by monitoring the time each vessel spends at each port, and matching that with the cargo type of that port and the load/offload rate. Other sources of cargo information for multi-cargo vessels include broker data on charter fixtures, bills of lading, vessel self-reporting, and personal communications with individual vessels, their owners, or operators.”
Rules-based logic determines whether each vessel is loaded or empty. According to the patent filing, “If a cargo vessel spends more than X number of hours at a certain export terminal, then the rules-based logic designates that vessel as ‘loaded’ when it departs that export terminal. If a cargo vessel spends more than X number of hours at a certain import terminal, then the rules-based logic designates that vessel as ‘empty’ when it departs that import terminal.”
Time spent at an export terminal is used by the system to determine how much cargo is loaded. “For example, if a crude oil tanker spends six hours at a crude oil export terminal with a 10,000 barrel per hour load rate, the rules-based logic calculates that 60,000 barrels of oil were likely loaded on that tanker,” the company explained in its patent filing, adding that the system’s calculations also take into account the properties of a cargo, such as the differing weights of various crude grades.
The height a ship is above the water is also utilised by the system to determine loaded cargo volumes. “Vessel height above water can be detected by satellite, land-based, sea-based, or air-based surveillance systems, including remote sensing or visual observations by humans (harbour masters or port agents), and web cams in ports or other locations,” it said.
Yet another challenge to providing a truly complete picture of global trade flows relates not to ship movements and cargo types, but rather, to cargo availability. What vessel interests want to know when setting a charter price is what cargos are actually available from shippers, and thus, what the real vessel demand is. Asked by Fairplay about this issue, Borgerson responded, “We’ve solved that, but I’m the CEO of a technology company that has been in stealth mode for nine years, so I’m not going to tell you how we solved it.”