A European airline operates a fleet of around 100 aircraft and offers both domestic and international flights. The airline is renowned for its punctuality and customer service, making it a favourite of business travellers. Its southern European focus has created a demand all year round from northern Europeans seeking escape from cold climates.
The airline has done relatively well during the pandemic compared to its competitors; it’s managed to cut down on routes in lesser demand, reduce staff, and generally slash costs without compromising the quality of their service.
But the pandemic has created new and different challenges for airlines. With months of difficulties with endless cancellations and restrictions, data is eroded with small and large gaps which make analysis and predictions difficult. Traditional methods and airline-focused IT systems are not able to look past the pandemic-infected data which has forced many pricing and revenue management teams to return to more manual methods. This is both costly and time consuming, and it prevents revenue management from taking full advantage of advanced pricing strategies.
The airline uses an industry-focused IT system that has a built-in demand forecasting mechanism. Although the Revenue Management has used the system-generated forecasts, they state that they have consistently lacked accuracy despite enormous amounts of data; something they accepted as “normal technological limitations”.
But months of cancelled flights due to the pandemic have left gaps in historical data, making the system-generated demand forecasts highly unreliable and creating headaches for the Revenue Management. As flights are picking up again, the airline has had no forecasts to rely on and has mostly been using the previous 7 days as the indicator of expected demand, as all other attempts to forecast accurately haven’t worked properly.
The airline approached SUMO Analytics to investigate if there was a way to forecast demand with higher accuracy and consistency than what they had achieved themselves.
The airline was particularly interested in having clarity for the 14 days prior to departure. After investigating the data along with the airline’s data science team, we customised multiple forecast models for different time frames, using pre-pandemic data. Multiple external variables are included to take into account anything that can impact demand, particularly different Covid-19 restrictions in different European countries. Real-time updates of booking data are constantly monitored and adjustments are applied to reduce potential forecast errors; i.e. this AI-driven forecast technique automatically adjusts forecasts for all departure dates and locations based on real-time live booking data.
We trained and set up the models in Azure, automatically pulling airline data from their systems and sending the forecasts back within seconds, which allows for constant adjustment of prices as demand changes.
This bridge from pre-pandemic data to the current situation along with customised modelling, looking at different time frames, generates much more accurate demand forecasts for airlines than system generated forecasts.
The pricing and revenue management team is now able to respond more quickly and, more importantly, more accurately.