Project Portfolio
Innovative Problem Solving
Sample Case Study: Solving a Difficult Revenue Loss Problem through a combination of text mining, sentiment analysis and statistical hypothesis testing
The Problem
A long-successful restaurant chain noticed a decline in same-store sales for a few of its locations, but couldn’t figure out the cause.
The Solution
Text and sentiment mining of Google and Yelp reviews revealed an erosion of customer satisfaction across a few key areas at the struggling locations which correlated with their decline in sales, as confirmed via statistical analysis.
Falling Restaurant Revenue Case Study Details
Problem Description
An Unexplained Downturn in Performance of a Long-Successful Brand
Olive Garden is a restaurant brand owned and operated by Darden Restaurants, Inc. As of 2008 same store sales had increased for Olive Garden for 55 consecutive quarters. Starting in 2009, however, same store sales for certain locations began to decline, and the Olive Garden brand had continuous erosion in the subsequent years. While a simple explanation might have been the great recession of 2008, this did not appear to explain the continued deterioration in same store sales numbers despite an improving economy post-recession.
A Data Driven Approach to Understand the Problem
Darden wanted to understand the key factors affecting Olive Garden performance using a data driven approach so that they could make more informed decisions and analyze potential risks.
Approach
Testing Multiple Hypotheses Using All Available Data
Chi-square testing and multiple linear regression were used to test a variety of hypotheses. Models made use of proprietary sales and operational data as well as geographical, demographic, and macro social and economic data.
Text mining and sentiment analysis were employed on a set of over 220,000 online reviews from Yelp and Google
Business domain and consumer behavioral pattern knowledge aided in guiding the analyses
Outcomes / Impact
A Decline in Service Quality for Certain Store Locations Over Time
The key to solving the mystery came from the text mining and sentiment analysis of the large dataset online reviews for store locations across the U.S. It turned out there was a statistically significant correlation between a decrease in same store sales and an increase in negative reviews that specifically mentioned terms pertaining to "service" and "employees“.
In addition to this crucial pattern, competition-related factors (and to a lesser extent, macro-economic factors) also contributed to the decline in same store sales for certain locations.
The Olive Garden chain of restaurants continues to operate and grow; it has expanded to 10 countries outside the U.S. as of 2024.
Falling Restaurant Revenue Case Study Images
Case Study - The Olive Garden Problem
Case Study - The Olive Garden Problem - Deliverable Detail 1
Case Study - The Olive Garden Problem - Deliverable Detail 2
More Examples of Problem Solving Projects
Second-Line Lung Cancer Treatment Decision Model
Healthcare / Biotechnology
Devised a pilot treatment decision model for non-small cell lung cancer patients requiring second-line treatment. The decision model combined survival probabilities and incremental cost data for each potential treatment pathway (including clinical trial therapies) and resulted in a shortlist of recommended treatment pathways, ordered by cost per QALY gained.
Novel Employee Engagement Analytical Model & Survey
Management Science / Human Resources
Designed a novel work engagement survey and analytical model based on a synthesis of three employee engagement theories from academic literature. This client project ignited a deep interest in this field of research for me. I retained IP ownership of the theoretical model (which I call the 'Four Lens Model') and hope to further develop and test it..
Customer-Facing Real Estate Ranking Algorithm, Survey & App
Real Estate
Developed a customer-facing survey and corresponding ranking algorithm and app to select and rank real estate properties in Mexico City based on weighted customer preferences, along with an internal-use companion app for analyzing and comparing real estate values and other metrics by neighborhood. Led the distributed data team tasked with building out a fully automated back-end system for the apps and integrated into existing company processes.
NIBRS Crime Data Query Tool App
Government Data / Research
Parsed and packaged millions of NIBRS data records; created a Shiny App for user-friendly data access and research purposes. The National Incident-Based Reporting System (NIBRS) is a nationwide (US) crime data collection and reporting system managed by the FBI. Anonymized data are released annually as sets of massive raw data files.
Retail Customer Segmentation Analysis & Targeted Strategy Recommendations
Consumer Goods / Retail
Deep-dived into historical data for a long-standing retailer in New Zealand transitioning to online sales while continuing to mail out physical catalogues at significant cost. Detailed statistical analyses led to recommendations for which catalogues to send, at what frequency, to which customer segments, in order to optimize conversion rate and order value.
Diagnostic Audit & Update Plan for Custom Builder's Revenue Forecast Model
Manufacturing / Construction
Executed a diagnostic audit and review of a monthly revenue forecast model used by a custom builder in Australia that tended towards over-estimating revenues in the near term. The deliverable was a detailed model update plan to correct the over-estimation problem and provide additional predictive accuracy.