At FLYR Hospitality, we are committed to providing our customers with the most accurate and actionable forecasts possible. Our forecasting model, developed by our industry-leading Data Science team, uses cutting-edge technology and algorithms to predict future hotel performance. This article aims to explain our forecasting process at a high level, compare it to the user forecasting workflow, and help you better understand how to interpret the accuracy of system forecasts.
How FLYR Creates Forecasts
FLYR employs advanced algorithms in a dynamic, evolving approach, leveraging diverse data sources to accurately forecast hotel performance metrics.
Key Points of Our Forecasting Process:
Data-Driven Approach: We use granular booking data, market comparisons, and event information to create a comprehensive view of future demand.
Sophisticated Algorithms: Our models are designed to adapt and learn, ensuring that they remain relevant and accurate over time. This includes using ensemble forecasting methods to combine the strengths of multiple predictive models.
Handling of Unconstrained Forecasts: We adapt our forecasts to account for both fixed and dynamic rates, applying traditional unconstrained models to fixed rates to gauge maximum potential demand. For dynamic rates, where unconstrained forecasts are not well defined, we adjust forecasts based on historical pricing and restriction behaviours. This approach is designed to provide a more accurate depiction of demand potential, tailored to the revenue management strategy in place.
Continuous Improvement: Our algorithms are regularly back-tested and updated based on their performance and new data, ensuring that our forecasts remain accurate and reliable.
Assessing Forecast Accuracy
FLYR’s nuanced approach to forecast accuracy distinguishes between system-generated and user-generated predictions. Rather than strictly adhering to traditional lead time benchmarks, we focus on perpetual model refinement through ongoing analysis against actual performance data. This method ensures our forecasts are as accurate and timely as possible, aiding your decision-making processes.
System vs. User Forecasts:
Granularity: FLYR generates detailed daily predictions for arrivals, length of stay, and rates, at the rate plan and inventory levels. This approach is distinct from user-generated forecasts, which concentrate on Units and ADR at higher levels of aggregation.
Adaptive Learning: Unlike static user forecasts, our system forecasts are dynamic, adjusting as new data becomes available. This adaptability can lead to differences in forecasted and actualized figures, especially further out from the stay date.
Accuracy Expectations: While users may target a specific revenue forecast accuracy threshold at a designated lead time, system forecasts aim for precision across all time frames, with the understanding that accuracy improves closer to the date of stay.
Interpreting Forecast Accuracy
When evaluating forecast accuracy, it's important to consider the context and the inherent uncertainty of forecasting. A deviation between system and user forecasts does not necessarily indicate an issue with the system model but may reflect the dynamic nature of the forecasting process and the different data and assumptions underlying each method.
Recommendations for Users:
Review Forecasts Regularly: As new data becomes available, forecasts will update. Regularly reviewing these forecasts can provide insights into trends and potential adjustments needed in strategy.
Consider the Lead Time: Understand that forecasts made further out from the stay date inherently carry more uncertainty. The accuracy of these forecasts should be interpreted with this in mind.
Leverage System Insights: Use the detailed granular data and insights provided by FLYR's forecasts to complement your own forecasting efforts. The system's ability to adapt to new data can help identify opportunities and risks not apparent in user-generated forecasts.
Conclusion
At FLYR Hospitality, we strive to provide the most accurate and actionable forecasts to support your decision-making process. By understanding the nuances of our forecasting approach and how it compares to user-generated forecasts, we hope you feel more equipped to utilize these insights effectively. As always, we are here to support you and welcome any feedback or questions on how we can further improve our forecasting capabilities to better meet your needs.