Enterprise Services Blog
Get the latest thought leadership and information about the role of Enterprise Services in an increasingly interconnected world at HP Communities.

Forecasting better Oil & Gas Results

I was doing my annual run through uptown Houston this past week with a few hundred of my closest friends when I took a breather and found myself going past some of Houston’s leading Oil & Gas firms.  As the EIA has just published its initial 2014 baseline and I have been in multiple discussions regarding how to improve forecast accuracies for production using analytics, I couldn’t help but spend some time in contemplation.    


If we look at forecasting completion times for a marathon and how they are used to place people in starting corrals, you quickly see contrasts between general guidelines and statistically-based forecasting.  For runners, the most common forecast metric is historical finishes.  Most runners will only improve so much from race to race and can be expected to be within a certain percentage of their past two years average. 


In fact, historical data is always important for any forecasting exercise.  The same is true for O&G industry forecasts, where we encounter a large number of variables.  However, we can expect a potential for very accurate forecasts based on geology, completions material and drilling and completion processes.  So, while the historical data in both situations is important in providing an accurate forecast for individual performance, enough information must be available to account for potential variances when making a forecast.  As a runner this might include illness or injury.  For oil production it might include type and amount of fluids and proppant.  To find meaning, these variables must be analyzed to determine potential causality, and not just casual correlation


For example, if we seek correlation only, we can discern that in the 10 years between 1983 and 1993 you could have used the production of butter in Bangladesh to predict how well the S&P 500 index would perform.  Or, if you prefer, you could ascertain that the stock market would rise in 2013, as the winner of the 2013 Super Bowl had its roots in the original National Football League.  Unfortunately, neither of these interesting correlations can be casually linked to the indicators they correlate to. 


If the exercise were focused on improving forecasts for my race finishing times, you would want to include consideration for injury, rate of recovery, amount of race conditioning, weather, type of shoe utilized, and weather-appropriate gear among others.  If the onion layers are pulled back for O&G, we know that certain factors will always impact production forecasts and pricing activities.  Any forecasting model needs to include the components and activities that are used to prepare and complete the activity regardless of the activity type.   


For instance, if I wanted to forecast potential pricing for heavy crude out of Venezuela to refine in Texas, I might need to factor in information such as pricing, demand for lighter sweet crudes, current and future production forecasts for heavy crude, recent shipment volumes, recent adherence to OPEC quotas, tanker availability and costs, utilization of competing refineries, cost of Saudi heavy crude, current percentage of hedge fund portfolios invested in oil, active tropical storms, and many additional factors such as current governmental policies and presidential health.  A series of algorithms and analyses would then be run to determine which factors showed causality compared to correlation – including potential leaders and laggards – to develop a hypothesis that can be tested against historical data. 


In the end, what we are seeking is a model that helps us anticipate potential production capability or pricing impacts so we can negotiate and secure the best available costs – based on their causality and impact on production or pricing.  This will hold true if we are working on models to forecast potential pricing and negotiation points for crude purchase, or point of sale pricing for gasoline at various locations throughout the our region.  I have worked with companies to use these types of capabilities to improve forecast accuracy from approximately 90% to over 98%, with subsequent improvements in their business performance.   


If we look at forecasting marathon completion times, I exceeded my expected time this year after having a limited training schedule in gorgeous weather.  Much better than the cold rain for the first 8 or 9 miles in last year’s run. 


Forecasting for the Oil &Gas business is the same process, but with many additional variables.  In order to do the best modeling, I advise clients about incorporating appropriate data, fit for purpose technology, and the tools that support accurate and sophisticated forecasting.  This enables an accurate method for supporting production and portfolio optimization, and the resulting improvement in business results.  Together, we build models, test models, then refine, reuse, and validate while making sure we can accommodate the unexpected, whether it is a cold downpour as you race or unrest in the Middle East. 


Capturing and utilizing the appropriate information to improve forecasting is just one of the ways Information Management & Analytics helps improve results for Oil & Gas companies.  You may find more about how HP supports Oil &Gas activities at hp.com/go/oilgas.



Tags: forecasting| Gas| Oil
Labels: Gas| Oil
Showing results for 
Search instead for 
Do you mean 
About the Author

Follow Us
The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation.