Baker Hughes accurately forecasted falling oil prices to enable better strategic business decisions
One of the world’s largest oilfield service providers enhanced the forecasting accuracy of its oil rig count model from 90 to 97.6 percent, predicted the impact of price changes on customer spending and demand for product lines (by leveraging 3,500 factors) and created a model that predicted the deflation of WTI crude oil prices in late 2014.
- Improve forecasts for key metrics, such as oil prices and rig counts, and economic activity
- Better understand the customer's decision-making process
- Identify best opportunities for capital investment and make better strategic decisions
IHS Markit Global Link Model, Economics & Country Risk, IHS Markit Enerdeq® and Energy Company & Transaction Research
- Predicted impact of price changes on customer spending and on demand for product lines, leveraging 3,500 different factors
- Enhanced forecasting accuracy in oil rig count model – from 90% to 97.6%
- Created highly accurate (8.5% error rate) crude oil price model that predicted the deflation in WTI crude oil prices at the end of 2014
- Enabled executive teams to make more informed decisions for optimal capital allocation – driving opportunity investments
Baker Hughes Accurately Forecasted Falling Oil Prices
As one of the world's largest oilfield services providers, Baker Hughes conducts operations in more than 80 countries – specializing in everything from drilling, evaluating formations, completions, production, and reservoir consulting. Baker Hughes' continued success depends upon the company's ability to understand myriad economic, strategic and competitive factors, particularly the impact of unexpected events, in order to minimize risks and maximize opportunities.
The Baker Hughes team leverages extensive resources to maintain a logical framework for tracking, analyzing, and predicting developments throughout global energy markets. They use sophisticated artificial intelligence, neural network, and machine-learning techniques to reduce close to 3,500 different factors to the ten core factors with the greatest potential impact on their business.
Baker Hughes researchers use extensive IHS Markit data, forecasts, and analytics as part of this process. For example, they use IHS Markit Enerdeq® and Energy Company & Transaction Research for data, such as well counts, mergers and acquisitions; where oil companies are purchasing acreage, and acreage valuations. Using the IHS Markit Global Link Model, Baker Hughes data scientists can analyze the impact of currency fluctuations, customer activity and hydrocarbon prices. By connecting and combining integrated information, analytics, and expertise, they were able to consider how these factors might influence individual customer decisions and allocation of capital.
Making Good Forecasts Even Better
Baker Hughes has been providing oil rig counts to the public since 1944, serving as an important barometer of industry activity and trends for the entire sector. The company gathers relevant data through field service personnel, who obtain the frontline data from routine visits to the various rigs, customers, contractors and other outside sources.1 The company's forecasts in some markets had an accuracy rate of 90 to 97.6 percent.
As Mike Daniel, Senior Manager for Market Intelligence for Baker Hughes said, "Our previous models gave us good predictions based on the best information available; but we wanted to determine what would happen if unforeseen events or factors emerged." For example, the team used the IHS Markit Global Link Model to understand what fiscal balance sheets, interest rates, and currency valuations would look like under economic stress.
Anton Gordon, Manager, Strategy and Corporate Development at Baker Hughes, and the sole developer and architect of the machine-learning applications, said, "Our oil rig count model is now very strong; it has an average of 97.6 percent accuracy. We also developed one of the most accurate crude oil price forecasts available, with an average error rate of only 8.5 percent. This helped us more accurately predict the recent drop in WTI crude prices at a time when many firms predicted the opposite, expecting crude prices to top $100 at the end of 2014."
We use predictive analytics models to glean patterns that we couldn't see before. We can take different factors and integrate them into one model and then develop a forecast that learns the patterns.
More widely, the team was able to leverage all of the "links" and "interconnection points" in the model to look at econometric data in a completely different way. "We use predictive models to glean patterns that we couldn't see before," said Gordon who designed the algorithms used. "We can take different factors and integrate them into one model and then develop a forecast that learns the patterns."
Those patterns translate into new forms of business intelligence. "For example," explained Gordon, "if you take the data from IHS Markit [Energy Company & Transaction Research] and start looking at where operators are buying acreage, we begin to learn the pattern that drives them to buy that particular acreage worldwide.. So, we say… if hydrocarbon prices go up, does the Asian market look more favorable versus the U.S. market? If prices fall to, for example, $50 in a U.S. market, what happens to activity in the U.S. versus the Euro Zone? Do we start seeing capital investments move out of the U.S. and go to Europe? Do markets like Latin America look more appealing due to the break-even prices that we're going to see there versus what we see in other markets? Again, it's more of a holistic approach. You can start looking at markets in a much more dynamic way. You can look at customers as discrete entities instead of thinking that, 'prices go up and activities are going to go up evenly' or 'prices go down then activity's going to go down evenly'."
Gaining Insights into Key Decision Factors
Baker Hughes' capabilities go far beyond forecasting hydrocarbon prices or rig counts. The company also uses leading-edge models to understand, in a holistic manner, how customers make decisions. Rather than looking at its customers as a group with activity that rises and falls evenly, Baker Hughes can view customers as discrete entities and consider where they are likely to allocate capital.
- Customers – How customers might spend at different price points
- Product lines – Which product lines might be affected as commodity prices change
- Overall global spend – Where global spending shifts after prices change
While Baker Hughes currently uses the IHS Markit Global Link Model to make predictions one to two years into the future, it is also testing to extend its forecasts out 5-10 years. Said Seema Santhakumar, Senior Manager for Global Market Analysis at Baker Hughes, "We look forward to providing guidance in terms of what prices and customer decisions will look like in multiple years—not just on an aggregate basis, but on a product-line basis too."