Technology fuels energy trading gains
AI-based trading platforms can drive transparency in LNG, commodities trading
Paris-headquartered Kpler has been shaking up the commodity trading market in recent years, using its proprietary AI-based software to track seaborne flows of liquefied natural gas, crude oil, refined products and dry bulk in real time around the globe.
By collating data from a wide range of sources and running it through advanced algorithms based on pattern analysis, machine learning and linear programming, Kpler can track the source and destination of individual cargoes, revealing hidden patterns and trends in the trading market which would otherwise go unnoticed.
Such is its popularity, Kpler is being increasingly adopted by trading houses, oil and gas majors, national oil companies, hedge funds, asset managers and investment banks wanting to track global supply and demand balances. It is also helping pinpoint trends and behaviours in the market to derive new trading strategies.
And it has a number of use cases: from determining whether Opec crude oil exports are over or under reported, to forecasting the volume of Qatari LNG exports destined for Europe, to tracking crude oil flows from a distressed market such as Iran or Venezuela.
It is single-handedly driving transparency in the rapidly evolving LNG spot market, a marked change on recent years when the LNG information landscape was essentially split into two, says CEO Francois Cazor.
"On the one side you had the operations and shipping angle that were following mainly a case by case story about vessels, with a lot of shipping data and logistics information. On the other side, you had trade flow analysis with a very macro level picture using figures mostly published by the countries' customs office with a time lag of two weeks to several months," he says.
"The market was trying to navigate the space between these two very different perspectives and the issue was that the figures would never reconcile—so you'd have the customs office saying 10 vessels discharged this month and the logistics would have seen 11 vessels, for example. If you add to this the complexities of the various countries and companies involved you get a big mess in terms of what's happening in this market. What we've done is to bring clarify to the data, bridging the gap between shipping and trading," adds Cazor.
In the past, traders also had to piece together much of the data themselves using a range of different tools. Kpler now does this legwork for them, making analysts more efficient by providing greater data precision and transparency.
Kpler's algorithms process satellite feeds, governmental databases and shipping registries, scanning, cleaning and structuring data from port authorities, ship agents, customs websites or brokers' reports, which are cross-referenced with market news. In essence, the service aggregates and connects the dots between a wide range of different datasets and presents it in a meaningful way, says Cazor.
"When we started out other providers were focused on delivering as many datasets as possible without taking much care about the quality. We took the opposite approach which was to say you don't need everything. A trader doesn't really care about all the logistic details, they just want to know who's buying and selling. The idea is to navigate this sea of data by giving them just what they want all in one place. The rest is available but on demand".
It means the service is constantly evolving in response to user needs—and data provision—with plans to add features which will enable the production, storage and consumption of molecules to be accurately and instantaneously tracked, as well as their seaborne transport.
For LNG trader Jean-Christian Heintz at Wideangle LNG, who was an early adopter of Kpler and has helped shape some of its features, the value is in its clear, user-friendly interface and the fact that it has teams on the ground in Houston, New York, London, Paris, Dubai and Singapore who are reactive and interactive.
"For the LNG market, it's putting more pressure on producers and buyers and making it easier for new players to understand and enter the markets," says Heinz. "In the past, it was a club and you had to get introduced to the right people and gather a lot of accurate information. Now if I'm in Bangladesh and I want to start a trading company you immediately have access to who's doing what with whom and can gain a lot of time".
"It's been playing a role where all the new generation of LNG buyers and producers have a totally different mindset now. They don't feel like they have to invest time in all these relationships," Heinz adds.
A core strength is that Kpler is open about its methodology, telling users where the data is sourced from—be it a port authority, a broker report, a ship agent report or a logistics company. This provides credibility and also makes it clear where there are gaps in the data. "That's made a huge difference because the industry was populated with a lot of poor ethics of data gathering where people were pretty much filling gaps," says Cazor.
The data is also based on physical reality. "We're showing what's happening in the real world. There is a simple thing which is physical reality. You cannot teleport a vessel. It has a limited speed, so sea conditions give a good idea of shipping capacity and what can realistically be moved from A to B in a few weeks. If a vessel takes a partial cargo, it's going to be quicker to load so we're going to see that a vessel stays a shorter period of time on the jetty. It's also going to be less heavy than if it's fully laden so we're not going to have the same kind of metrics".
At the same time, there are limits to Kpler's service and what it can do. "We're never going to be better than Shell and to be able to tell them what they're doing, but we can be quite good at telling Shell what BP is doing and vice versa".
Nonetheless, the service is growing in capability all the time, building on 10 years of historical data, and counting, to compile an ever-deepening pattern of trading behaviour with which to predict trade flows. The rising tide of data from sensors, nano-satellites and other monitoring technology means there is no shortage of information for Kpler to collate and analyse.
"The more data we collect the more we can predict, so every day the system gets smarter in terms of analysing the patterns, the correlations between all the different categories," says Cazor.
"The data landscape, logistics data, commercial data, weather data, maritime conditions and so forth are creating several arrays of datasets. With machine learning technology behind it, we can predict or at least with a certain confidence rate what might happen in the next weeks or months".
A central appeal of Kpler is that, by taking a data-centric approach, its service is unbiased. At times this means Cazor has had to reject requests by some users to remove certain information or stop tracking their operations. "We're not bullish or bearish we just give data and if you make or lose money that's not our business," he says, adding that limiting the ability to make money is a natural evolution.
"All markets take the same path, at some point you get more transparency, buyers are going to be more educated, sellers are going to be more challenging or vice versa and naturally the margins will shrink but it will come with volumes in trading so all in all everyone's happy," says Cazor.