Revenue Managers and Demand Forecast Accuracy
Revenue managers must constantly be on the lookout: they must monitor market trends and competition, they analyse new systems and different variables, they evaluate information and transform all these data into specific strategies to be employed by the hotel. However, Revenue Managers also face some obstacles that do not allow their strategies to reach their full potential. What we will discuss today is one of those obstacles: demand forecast accuracy.
Revenue Management's goal - maximizing revenue and, ultimately, profitability - is reached through demand-management decisions such as forecasting demand and its features while applying inventory and price control to manage it. However, the accuracy of this forecast is heavily impacted by reservation cancellations.
Various studies highlight that cancellations can affect somewhere between 8% to 26% of the reservations received by hotels, so a precise forecast for reservation cancellations and reservations in general is vital because, in the end, it is the hotels that incur the cost of having empty rooms when guests cancel or do not show-up.
Hotels use overbooking and restrictive cancellation policies as tools to minimize that from happening, however both these techniques might have an adverse consequence on hotel performance and reputation, if not managed appropriately: a guest will be displeased to know he just has been overbooked which can originate complaints and, in turn, will create negative impact on the hotel's reputation not to mention the costs associated with placing the guest in a different hotel. Restrictive cancellation policies, for example, might result in a reduction in the amount of reservations and subsequent reduction in revenue. All these have significant impact on the guest, the team and business.
I believe a potential solution for this challenge can be the implementation of a machine learning algorithm to create a model that can forecast the bookings that are most likely to cancel.
Machine learning algorithms have shown, in recent studies, accuracy and precision levels of more than 80% in a period of almost 2 years.
The algorithm can take variables that are known to be determining factors when it comes to cancellations, such as lead time, country of residence and booking channel, combined with historical data and it will give the probability of cancelation for each individual booking.
By knowing who is likely to cancel, I am sure hotels can act to try to avoid cancellations by, for instance, contacting the guest, but it can also help revenue managers to fine-tune their overbooking and cancellation policies strategies, as well as pricing and inventory allocation decisions. With this information I believe hotels can present different cancellation policies to guests forecasted to be less probable to cancel than to those that are anticipated to be more prone to cancel.
However, machine learning models need to be applied with caution. Practitioners need to take into account the expenses related with gathering, storage, and treating information, the time necessary to process great amounts of data and the time consumed in preparing data and modelling for each property. Consequently, the use of these models calls for considerate evaluation of the associated pros and cons.
Since we are in the early stages of its adoption, there is still much to be learned about its capability and use, but there is definitely potential. Unfortunately, the high costs for implementing this system makes it hard for every player to apply, however, I firmly believe that as technology evolves, machine learning will become the norm.
References
- Antonio, N., de Almeida, A. and Nunes, L., 2019. Big Data in Hotel Revenue Management: Exploring Cancellation Drivers to Gain Insights Into Booking Cancellation Behavior. Cornell Hospitality Quarterly, 60(4), pp.298-319.
- Falk, M. and Vieru, M., 2018. Modelling the cancellation behaviour of hotel guests. International Journal of Contemporary Hospitality Management, 30(10), pp.3100-3116.
