S&P Global Platts Taps Kensho's AI to Solve Commodities' Big Data Woes

After acquiring AI specialist Kensho Technologies almost two years ago, S&P Global is keen to leverage Kensho’s skillsets across its business silos.

S&P Global Market Intelligence, which provides Kensho with market data, began by using Kensho to build out its Global Marketplace. This will serve as a “storefront” for the data provider. Now the company is expanding Kensho’s focus to S&P Global Platts, a source of benchmark price assessments for the commodities market.

Platts is using Kensho’s big data manipulation capabilities internally to structure and source large volumes of disparate data, while utilizing Kensho’s data visualization, charting, and natural language processing (NLP) capabilities to improve external, client-facing functions.

Silvina Aldeco-Martinez, S&P Global Platts chief product officer, took up the role about 18 months ago. She was struck then by the wide gap in technology and data capabilities available for commodities versus other markets. 

“The level of comfort and confidence using deep data and quantitative tools in the commodities market is several years behind where we are in equities, or where fixed income is,” Aldeco-Martinez says. “So [I thought], How can I bring best practices into the commodities space? And that is what we have coined ‘digitalizing commodities markets’.”

One of the key processes that Kensho has optimized is Market-on-Close (MoC), which assesses prices for crude oil, petroleum products, and swaps, among other investments. Currently, market participants submit offers and transactions to Platts’ editors. These are published in near real time throughout the day until market close. Following the close, editors examine the data gathered throughout the day, conduct their analyses, and develop price assessments that reflect an end-of-day value.

Of the roughly two-hour MoC process, an editor spends about an hour and a half aggregating, formatting, and double-checking data that comes in through a variety of sources, such as APIs and broker calls. Kensho is stepping in to connect to those APIs and to use machine learning to bear the brunt of those tedious processes. As a result, Platts can now deliver key benchmarks to market participants up to 25% faster, Aldeco-Martinez says.

On the visualization front, Kensho NLP automatically structures numerical data that is constantly changing, and presents that information in a way that is more easily consumable for humans. Traders are already used to quickly interpreting their screens’ series of rapid flashes that signal bids, sells, and asks; newer features augment the user experiences of some of Platts’ other customers, who are often in risk management oversight roles and who have to monitor multiple markets at once. As these flashes appear in real time, they’ll appear with text that indicates sellers’ names and offers, subsequent bids, quantities, and buyers.

As those offers and bids come together, charts will be generated in real time as the market unfolds throughout the day. Companies will be color-coded and linked by lines of varying thickness to illustrate larger and smaller stakeholders, and to show the dynamics at play between them.

“So just by looking at that spider web, you start to understand who’ the market-maker is for today, who they are buying or selling to, and who they are buying or selling from,” Aldeco-Martinez says.

Demand for increased accessibility will persist, she says. A frequent request from customers is for the ability to view the market and their screens on mobile devices, and Aldeco-Martinez says the company is embracing this, particularly as the commodities market strives to match its equities and fixed-income counterparts in technology, and as the market itself becomes more complex.

“We are a unique company that can walk you through what’s happening based on industry fundamentals, all the way to what the market is doing right now, and into the future through our analytics forecasts—and we do it across commodities,” Aldeco-Martinez says. “More and more, understanding that cross-commodity energy balance or equation is so critical because if, for example, in the agricultural world the production of sugar is deriving or creating more growth in bioethanol, that could start competing with traditional fuels.

“So now you start to see why it’s important to get a good read of what’s happening in the sugar crop in order to understand potential competing alternatives to some of the refined oil products.”