Southwest has a fleet of 700+ airplanes serving 101 destinations. Each one needs the correct amount of fuel, keeping in mind operational and financial bias. For example, the company needs to purchase and store enough fuel for scheduled flights and unscheduled events that require additional fuel. The company should also buy in bulk at low cost while considering storage costs and capacity at each airport. Keeping planes fueled and on time is imperative for customer service and loyalty, so forecasting is vital.
Southwest now uses Alteryx to power its fuel consumption forecasting. Southwest’s financial planning and analysis team previously used Excel to conduct a simple forecasting method for every airport each month, which was time consuming, took up to three days, and opened possibilities for error.
Using Alteryx, the team created an advanced data model that has improved the accuracy of fuel consumption forecasting at the airport and system levels and helps fuel inventory management. The consumption forecast provides a 12-month horizon for the fuel supply chain department, improving accuracy by incorporating factors that can affect fuel consumption (estimated number of trips, month, fuel price, etc.), identifying problematic airports/forecasting months, and preparing decision makers against unexpected conditions.
The team is experimenting with machine learning and advanced statistical modeling to loop together a series of predictive models to find the right model to apply during different seasons at different airports. For example, in January, when weather conditions vary, a non-linear regression model may work best to account for weather variance. In June, exponential smoothing with seasonality may be better.
The forecasting accuracy of scheduled flights improved approximately 12 percent, which can result in millions in fuel cost avoidance. On a personal level, the analysts finally have their Sundays and nights back: the new process reduced implementation time from three days to just five minutes.