Prioritized Recommendations for Hotel Properties - Helping hotel property managers prioritize actions to efficiently drive guest satisfaction

Description:

This tool is being presently used by a major multinational hospitality company.

Business Objective - Hotel chain with more than 6,000 properties around the world wanted to help property managers improve guest satisfaction. Their descriptive dashboards, summarizing a database of guest surveys weren’t enough. Low survey scores were compared for properties of the same brand and continent. We found that survey questions with poor scores were not necessarily driving a reduction in an overall Satisfaction KPI, and that there was great variety in properties with traditional segments.

Innovative techniques used -

1. Random Forest was used to model the relationship between specific survey questions and the Satisfaction KPI in the context of each property’s unique characteristics. Sensitivity analysis was used to predict the Satisfaction KPI improvement from a 1-point improvement for each question.
2. Missing values for unanswered and unasked questions were imputed Deloitte’s Matrix Completion Engine and business rules.
3. A 14-dimensional “Property DNA” was developed from over 80 variables, many of which were categorical, using PCA on groups of related variables. Dissimilarity of properties could then be computed using a simple distance calculation.
4. ROI was used to prioritize which survey question scores to improve, combining the sensitivity coefficients and an estimate of the difficulty of making a 1-point improvement.

The tool helps property managers focus improvement efforts on areas that will most likely increase guest satisfaction. It gives property managers a prioritized list of scores to improve, customized to the individual property’s situation; chosen to provide the best ROI for improving the property’s Satisfaction KPI. It also identifies, for the first time, other properties that are most similar, allowing managers to see a meaningful comparison of their property to others.

Prerequisites:

• Basic understanding of Trees (Random Forest)
  • Clustering

Content URLs:

https://drive.google.com/file/d/1-zz4ju2y75Gl50NqNw9XsczitTQIU0T8/view?usp=sharing

Speaker Info:

Ritika Kapoor - Data scientist with more than 3 years of rich industry experience. Skilled in Deep Learning and advanced statistical learning methods and have leveraged the same to make operational performance improvements for clients. Skill set includes Computer Vision, Metaheuristics, Recommender Systems, Retail Analytics, Sales Forecasting, Optimization, Sparse Data Handling, Banking Analytics.

Monisha Ahuja - Currently working as a Data Scientist with Deloitte USI Consulting with 1.5 years of experience in predictive modelling space. Have expertise in sales forecasting and recommendation system.