Client analysis found they would have saved $7 million dollars by hedging feedstock costs based on ChAI’s forecasts.
One of Asia’s largest integrated petrochemical companies wanted additional support to optimise their risk management strategies in the second half of 2019. The client wondered whether it could use the predictions created by ChAI’s machine learning algorithms as alternative information to consider better hedge positions depending on ChAI’s market predictions.
The client needed to lock in the margin between the price it paid for feedstock and the sales price to the customer. Since ChAI started monitoring oil prices back in the middle of 2019, Naphtha - the feedstock that the client needed to purchase - had been exceptionally volatile. Prices rallied to nearly $600 per tonne by early 2020, before falling over 70% in value by the middle of March 2020. Brent crude oil, which is a major foundation for Naphtha forecasting, was also extremely volatile during this period. The client regularly hedged its purchases, opting to hedge more as prices rose and less when they fell. By hedging more as prices rose, it would achieve more protection against higher prices. But if the client hedged for the month as prices fell, it would post a financial loss.
The client was able to use ChAI as an extra input to gauge when, and how much, to hedge. ChAI deploys artificial intelligence to predict short-term price trends, making it an ideal way for the client to manage its hedge book. When ChAI predicted Naphtha prices higher than the client’s estimate, the client had the option of increasing the volumes it hedged to buy greater protection. Hedge managers make similar decisions on the scale of protection they need every day. Using ChAI gave the client an additional, timely input, based upon the latest machine learning techniques that process thousands of relevant and often hard-to-source data points.
Using ChAI data over the period between Q3 2019 and April 2020, the client would have theoretically saved millions of dollars by optimising its hedges on Naphtha. In four out of the five months that were trialled for Naphtha, ChAI’s predicted price was closer to the ultimate price than the client’s, and demonstrated a directional accuracy of 80% in the price change predicted, as well as a 99.7% mean absolute percentage accuracy.1
The client would have booked millions of dollars saving on its hedge positions as well as additional savings in feed management by following ChAI’s market advice. By looking at ChAI’s price predictions for Naphtha, the client could decide when and how much to hedge. ChAI helped its client make more informed decisions around its risk management requirements, leading to potentially higher margins and better budgetary control.
¹ Mean absolute percentage error (MAPE) is a metric which interprets the distance between the forecasted price and the actual price obtained on the day of prediction, scaled by the price itself. We define the more interpretable Mean Absolute Percentage accuracy to be equal to 1 - MAPE here.