Maximizing Potential Revenue of Global Hotels and Resorts

As part of the Saunder's Business Analytics Case Competition hosted by the Rochester Institute of Technology, Team Robolytics, comprising talented students from Simon Business School, embarked on a strategic project aimed at maximizing the potential revenue for Global Hotels and Resorts in Lisbon, Portugal. This initiative leveraged advanced analytics and predictive modeling to uncover actionable insights for enhanced revenue management. The team secured second place in the competition.

Team Robolytics Members:

Saheel Chowdhury (Nationality: Bangladesh)
Moemedi Wazzza Rakhudu (me) (Nationality: Botswana)
Aditi Jain (Nationality: India)
Joseph Pellumbi (Nationality: USA/Albania)

Together, this diverse team combined their expertise to tackle the intricate challenges presented by the hotel's existing data, aiming to formulate a robust strategy to significantly boost the hotel's revenue streams through meticulous data analysis and strategic planning. Their efforts highlight the importance of interdisciplinary approaches in addressing business analytics challenges in a competitive setting.

DOWNLOAD SLIDE DECK

Objectives and Methodology

The main objectives were to assess the impact of room upgrades and last-minute cancellations on revenue, and to predict the likelihood of free upgrades. To achieve this, the team employed an advanced methodology that included data cleaning, augmentation, and segmentation, using a combination of Python, R, and Tableau for thorough data manipulation and analysis.

Key Findings

The analysis revealed significant insights:

Revenue from Hotel Stays: The hotel’s revenue from stays was substantial, but opportunities for increasing it were identified in the patterns of upgrades and cancellations.
Impact of Upgrades: Upgrades often resulted in lost revenue due to the disparity between the actual room price and the price paid after upgrades.
Cancellations: Last-minute cancellations had a pronounced negative impact, especially from bookings originating within Portugal and nearby countries.

Data Cleaning and Augmentation

For their project with Global Hotels and Resorts, Team Robolytics undertook meticulous data cleaning and augmentation efforts to ensure the quality and usability of their dataset for analysis. Here’s a detailed look at the steps involved in their data preparation process:

1. Adjusting Metadata and Handling Missing Values

Initially, the team adjusted the dataset's metadata to align with their analytical requirements. This included redefining data types and labels for clearer interpretation and easier manipulation. They also tackled missing values, a common issue in large datasets, by identifying and imputing them where appropriate or removing entries that could not provide reliable information.

2. Data Cleaning

The team's data cleaning process was extensive, involving:

Removal of Outliers: Outliers that did not fit the general booking patterns or pricing models were removed to prevent skewing the results.
Consistency Checks: Ensuring consistency across data entries, especially in terms of formatting dates, currency, and categorization of room types,

3. Data Augmentation

Data augmentation involved enhancing the dataset with additional calculated fields to provide more depth to the analysis:

Total Stay: Calculated as the sum of weekday and weekend stays per booking, providing a comprehensive view of guest duration.
Total Guests: Summed up all guest types per booking, including adults, children, and babies, to get a complete count of occupants.
Upgraded: A binary field added to denote whether a booking was upgraded from the initially reserved room type, crucial for analyzing the impact of upgrades.
Booking Revenue: Computed as the product of the Average Daily Rate (ADR) and Total Stay, giving a direct measure of revenue from each booking.
Actual Room Price: A critical metric derived by calculating the average ADR per room type per season, providing a basis for assessing the financial impact of room upgrades.
Potential Revenue Lost: Calculated by subtracting the ADR from the Actual Room Price and multiplying by the Total Stay, highlighting revenue lost due to discrepancies in room pricing.

4. Data Segmentation

To tailor strategies based on varying conditions, the data was segmented by different periods:

Yearly Segments: The dataset was divided into the years available in the dataset, allowing year-over-year comparisons and trends.
Seasonal Analysis: Further breaking down the data into peak and off-peak seasons provided insights into how booking patterns and pricing strategies could be adjusted seasonally.

5. Cancel Lead Time

This newly introduced metric indicated the number of days before a booking was canceled. It was particularly useful for analyzing last-minute cancellations and their impact on revenue, allowing the team to propose targeted cancellation policies.

Through these detailed data cleaning and augmentation efforts, Team Robolytics ensured that their dataset was not only robust and comprehensive but also tailored to extract the most meaningful insights for revenue optimization strategies. This foundational work was critical in enabling the accurate predictive analyses and strategic recommendations that followed.

Predictive Analysis

In their project for Global Hotels and Resorts, Team Robolytics deployed several predictive models to address key challenges related to revenue management. These models focused on predicting the probability of room upgrades and estimating the subsequent revenue impact of these upgrades and cancellations. Here’s a closer look at the predictive models used and the insights they provided:

Predictive Models Used

Random Forest: This model was chosen for its effectiveness in handling large datasets with complex structures. It’s particularly good at capturing nonlinear relationships between variables which are common in hospitality data.
Generalized Linear Regression (GLR): GLR was used to provide a statistical framework for estimating the financial impact of upgrades, helping to linearly correlate features like booking lead times, room types, and stay duration with revenue loss.
Extreme Gradient Boosting (XGBoost): Known for its precision and efficiency, XGBoost was utilized to improve predictions related to upgrade probabilities and their financial outcomes.

Core Predictions and Insights

Probability of Free Upgrades: The models estimated a low chance of free upgrades, ranging around 2.68% to 2.87%. This low probability indicates that free upgrades are not a frequent occurrence but could still impact revenue significantly when they occur.
Revenue Loss Per Booking Due to Free Upgrades: The estimated average revenue loss per booking due to free upgrades was around $149. This figure highlights the potential revenue leakage that could accrue if upgrade practices are not managed properly.

Model Outputs and Interpretations

Random Forest: Predicted a 2.68% chance of free upgrades with a revenue loss per booking of $149.7. Confidence intervals provided by the model showed that revenue loss could vary between $145.2 and $149.7, offering a reliable estimate with minimal variance.
GLR: Showed a slightly higher probability of upgrades at 2.87%, with a consistent revenue loss prediction of $147.7 per booking. The tight confidence interval around this estimate underscores the model’s accuracy.
XGBoost: This model gave a similar upgrade probability at 2.78% and estimated the revenue loss at $149.8 per booking. The model's confidence interval suggested that the actual loss per booking would likely fall between $144.0 and $149.8, giving the team a precise range for financial planning.

Statistical Measures and Reliability

Log Loss: Used to measure the accuracy of the classification models, with lower values indicating better model performance. This metric helped the team fine-tune their models for optimal prediction reliability.
Mean Squared Error (MSE): This measure helped assess the variance in the models' predictions, ensuring that the predictions were stable across different data samples.

The predictive analysis performed by Team Robolytics provided robust insights that enabled Global Hotels and Resorts to understand the nuanced impacts of room upgrades and cancellations on revenue. These predictions formed the backbone of their strategic recommendations, ensuring that each suggestion was backed by solid data-driven evidence.

Insights & Interpretations

After developing and deploying their predictive models, Team Robolytics conducted further analysis to deepen their understanding of the factors influencing revenue loss and potential gain at Global Hotels and Resorts. This analysis focused on several key areas:

Market Segment Performance: The team analyzed the revenue contributions from different booking channels. They discovered that Online Travel Agents (TAs) generated the highest revenue, indicating that strengthening partnerships with these agents could be a lucrative strategy.
Geographical Impact on Cancellations: The analysis revealed that a significant number of last-minute cancellations came from Portugal and neighboring European countries. This insight was crucial for tailoring cancellation policies to minimize revenue losses from these high-risk areas.
Customer Segmentation: By examining the booking habits and loyalty of different customer segments, the team identified that corporate customers were the most frequent repeaters. This highlighted an opportunity to develop targeted marketing strategies to retain and expand this customer base.
Upgrade Analysis: The team also focused on the incidence of room upgrades, noting that Online TAs were responsible for the majority of free upgrades. This finding prompted a reevaluation of upgrade policies to prevent revenue dilution.

Recommendations

Based on their comprehensive analysis, Team Robolytics proposed several strategic recommendations to optimize revenue management for Global Hotels and Resorts:

Implement Tiered Cancellation Fees: To address the high incidence of last-minute cancellations, especially from nearby countries, the team suggested implementing tiered cancellation fees. These fees would increase as the arrival date approaches, discouraging last-minute changes and securing revenue.
Restrict Free Upgrades During Peak Seasons: The team recommended setting caps on the number of free upgrades allowed during peak seasons. To propose this developed a tool that would allow them to dynamically calculated the number of free upgrades that can be given out without incurring a loss. This would help maintain room availability at premium rates and ensure that upgrades do not undercut potential revenue.
Loyalty Programs for Corporate Clients: Given the high repeat booking rate among corporate clients, introducing loyalty programs tailored to this segment could enhance customer retention and increase long-term revenue. These programs might include flexible booking options, express check-in/out, and other perks that appeal to business travelers.

Special Packages for High-End Rooms: To leverage the popularity of deluxe rooms, the team suggested developing special packages that combine luxury accommodation with unique experiences, such as cultural tours, wellness retreats, or gourmet dining. These packages would attract a more affluent clientele and enhance the perceived value of staying at the hotel.
Dynamic Pricing Strategies: Employing dynamic pricing based on demand, booking lead time, and customer segmentation could help optimize room rates and maximize revenue across different market conditions.

These recommendations were designed to build on the predictive insights provided by the team’s models, offering a multi-faceted approach to revenue management that could adapt to changing market dynamics and customer behaviors. By integrating these strategies, Global Hotels and Resorts could not only address immediate revenue challenges but also position itself for sustainable growth in the competitive hospitality market.