View on desktop for the best experience

Hotel Cancellation Predictor (PowerBI)

Running a hotel is never simple, especially when a large number of guests change their plans at the last minute. Cancellations may seem like a normal part of the business, but when nearly one out of every three bookings disappears, the impact on revenue and planning becomes impossible to ignore. Rooms sit empty, forecasts lose accuracy, and staff are left scrambling to adjust.

This challenge inspired me to build a Hotel Cancellation Risk Predictor. Using data from more than 36,000 bookings, I explored the patterns that drive guests to cancel. The model looks at factors such as how far in advance a booking is made, how long the stay will be, the price of the room, and the level of guest engagement. By analysing these signals, it can predict the likelihood of cancellation with meaningful accuracy and, more importantly, highlight the strategies hotels can use to protect revenue and build stronger guest commitment.

Download PDF

* This file may display shadows in some viewers due to transparency effects; for best results, please use Adobe Acrobat Reader.

Hotel Cancellation Risk Overview

Methodology

The prediction uses a statistically calibrated model based on historical booking patterns to project cancellation risk. It helps evaluate policy options, pricing strategies, and overbooking thresholds.

- Predicted by the following factors: Lead Time, Nights of Stay, Room Price, Engagement Score, Credit Score, and Month of Arrival.

- Engagement Score = Special Request + Required Parking Space + Repeated Guest (Y/N)

- Credit Score = Previous Completed Bookings - Previous Cancelled Bookings

About the Study

This project analyses 36,000+ bookings at a hotel to identify guests most likely to cancel.

By modelling behavioural signals, we can predict risk at booking time and tailor strategies to reduce cancellations (at 72% accuracy) for real-time decision-making.