Cloud ML – Forecasting Booking Pattern

PROJECT BACKGROUND:

Our client, a global resort operator was seeking a solution to improve inventory & pricing management by forecasting occupancy rates according to historical booking patterns thus optimizing Average Daily Rate (ADR) and Average Occupancy Rates (AOR).

THE CHALLENGE

Our client retained several years of booking data, however, Revenue Managers found it challenging to assess & identify patterns in order to better predict and manage inventory & pricing. They were using a statistical methods, tools & calculations which were time consuming and often provided questionable accuracy and resulted in lost revenue opportunities. Identifying the booking patterns / behavior across multiples segments, resorts and revenue streams presents many challenges:

• Multiple inter-related revenue streams while managing demand & allocation.
• Years long of data across multiple seasons / resorts.
• Identifying correlations among different segments & how one affects the other.
• Importance of seasonality & external factors such as events, weather, etc.

A Scientific approach is needed to first identify the booking patterns across multiple segments, resorts and revenue streams. Then to utilize these patterns to forecast demand & manage pricing effectively. More importantly, the solution should be scalable, cost effective & implemented within a short timeframe.

OUR SOLUTION:

Vertisystem assisted our client in the design & development of a scalable, automated cloud-based Machine Learning solution which forecasts the demand & booking pattern leveraging the client’s existing GCP Environment & Looker. Vertisystem contributed to solution architecture, feature engineering, data preparation, batch prediction, data refresh automation and model evaluation/retraining processes. The complete solution provided a model using Cloud AutoML. In collaboration with the business users, Vertisystem developed Looker models, reports & dashboards which interact with existing revenue data & provides key insights using the forecasted data.

BUSINESS RESULTS:

• Revenue gains by effective management of pricing using the forecast data.
• Greater confidence in results due to leveraging a more scientific approach which reflects changing behaviors in travel situations like pandemic and other anomalies.
• Improved management of demand, allocation & pricing across different segments, properties, revenue streams.
• Visualizations to provide insights to business managers & users to plan effectively.
• Improved insights on expected booking pattern with respect to seasonality & shift between revenue streams.
• Ability to accurately forecast demand & provide booking pattern predictions.
• Scalable solution to support resorts globally with less work.
• Implementation completed in quick time by leveraging available solution opportunities with lesser cost.
• Fully automated data ingestion & management resulting in reduced man cost & managing costs.

Confidential – Vertisystem, Ltd.