How much oil will India use this month? It is a question that is both important, because of the country’s emergence as the world’s largest source of new demand for crude, and difficult to answer accurately. Indian oil consumption data are published with a time lag and initial estimates are subject to significant revisions.
, a Paris-based start-up, intends to find a solution to that problem. Its aim is to apply advanced analytic techniques to energy data that will allow more accurate assessments and forecasts.
Antoine Rostand, founder and chief executive of Kayrros, says the plummeting cost of computing power can revolutionise how energy traders and investors use information. “We can do analysis of text and images now that would not have been possible even a couple of years ago,” he says. “We can do in seconds what would have taken months.”
The search for ways to harness this potential has reunited two veterans of Schlumberger, the world’s largest oil services group.
Over a couple of decades, the balance of power in the global oil business has shifted from production companies such as , and towards service companies, with Schlumberger at the forefront. It invests a much greater proportion of its revenues in research and development than those companies and has come to dominate many of the industry’s critical technologies.
Andrew Gould, now an investor and the chairman of Kayrros’s advisory board, was ’s CEO in 2003-11. Mr Rostand began his career there as a field engineer, before working as a consultant and executive in the oil and IT industries, including a spell as CEO of Electronic Data Systems in France. In 2004, Mr Gould hired him to set up Schlumberger’s consulting division, which became one of the most admired advisory groups in the industry. When it was sold to Accenture last year, Mr Rostand decided to strike out on his own.
Schlumberger’s technological leadership and global expansion give Mr Gould a good claim to be considered the most successful British oil executive of his generation. After Schlumberger, where he spent 37 years, he had a second act as chairman of , the British oil and gas production company. He completed its sale to Shell at the generous price of $51bn in February this year.
Now 69, he says that was his last leadership role at a big public company. He was thinking about going into academia but was also interested in helping a start-up — and accepted the opportunity to back Mr Rostand a second time.
“I’ve always loved R&D and in Schlumberger I had a billion dollars of R&D [a year] to play with. And I’ve always loved the high-risk end of it,” he says, speaking over breakfast at the Four Seasons hotel in New York. “This is not pure R&D but the thought process is very similar.”
Kayrros has 15 employees, mostly data scientists, mathematicians and petroleum engineers, and aims to reach 35-40 by the first quarter of next year.
With the oil industry again in the doldrums, signing up petroleum engineers has been straightforward. Hiring world-class data scientists is trickier but Mr Rostand says his management team is a powerful draw. Laurent El Ghaoui, a professor at the University of California, Berkeley, is one of his founding partners: “When they look at our management, they know they will learn and progress,” Mr Rostand says.
He has put his head office in Paris, where there is an excellent pool of mathematical and engineering talent to draw on. The French government has also provided backing in the shape of a loan from the state investment bank that is available to innovative companies.
Kayrros — from Kairos, the Greek personification of opportunity — also has offices in Berkeley and New York, and engineers working remotely from London. Modern communications make it possible for the team to work together. They communicate on Slack, use Github for their software development and Asana for project management.
As a mentor and figurehead for this cerebral group of researchers, many at the beginning of their careers, Mr Gould, with his thoughtful, donnish manner, seems a perfect fit.
He does not have his own office at the company, but is available as an adviser and guide. “I am internally the person with the most senior knowledge of the oil and gas industry, so I do serve in an advisory role to the team when they come with ideas,” he says.
His other great asset to Kayrros is the contacts built up in a four-decade career. Because Schlumberger operates as a service provider to national oil companies and does not seek to take a share of a country’s production — unlike Exxon, BP or Shell — it is welcome in places such as Russia and Saudi Arabia where other western groups face difficulties. Mr Gould can open doors.
Forecasts of energy markets are notoriously unreliable. The key events of recent years, including the US shale boom and the plunging cost of solar and wind power, were largely unforeseen. Poor data quality often makes it difficult to know what is happening now, let alone make predictions. The penalties for using unreliable information, and the rewards for accuracy, can be huge.
The key to improving the quality of data and forecasts, Mr Rostand says, is machine learning: using computers that can adapt without being explicitly programmed.
Applying machine learning to large, complex data sets — both traditional sources such as official statistics and news stories, and newer ones such as satellite images and traffic reports — makes it possible to identify patterns and derive forecasts of energy supply and demand much faster than with previous techniques.
Traders that have reliable information unavailable to others have an edge in knowing when to buy and sell.
Using machine learning to analyse market data is a hot field. Eric Schmidt, chairman of , and other billionaires are backers of a hedge fund called CargoMetrics, which uses similar analysis of to find profitmaking opportunities in commodity markets.
Kayrros will sell its analysis to traders, fund managers and companies, rather than trading or investing for itself, but Mr Rostand argues that CargoMetrics is an important “proof of concept” that shows the power of the technology.
Although Kayrros has no fully finished products yet, it is signing up clients. “There are one or two who are prepared to subscribe on a fairly limited basis, because they don’t know exactly what products are going to be available. And there’s a huge interest among everybody in beta-testing the results,” Mr Gould says.
The investors so far include friends of Mr Rostand, two members of the original Schlumberger family, and Mr Gould, raising a total $1.2m. A second financing round is looming, probably late this year or in early 2017, but the company will need to be able to demonstrate some products first.
“It’s an ambitious project, but if he [Mr Rostand] manages just one or two really insightful analyses of data from sources that the industry didn’t traditionally use, it could have a tremendous amount of value,” Mr Gould says.
His new venture: “The ultimate aim is to use big data and machine learning to try to get to sources of data that have not traditionally been used in the oil and gas industry, or have been used but only partially.”
Running a public company: “I would not do another public company, that’s for sure. It was painful enough doing one.”
Oil prices: “When you look at the figures for the investment that’s been cut, the longer this goes on, the more dangerous it gets . . . There are still people who say ‘Oh, the shale guys are going to come back at $45 [per barrel]’. I doubt it.”