Hotel Performance FINAL

Georgetown Data Analytics

Hotel Performance Drivers

Ashley Loyd, John Cannon, and Monika Adamczyk

Fall 2014 – Spring 2015

Business Problem

Client:

Hotel Managers & Hotel Investors

Business Problem:

Recent figures show decreasing RevPAR in the Houston hotel market. Investors are worried and suspect that this is because of oil industry.

Hotel Performance: RevPAR

Revenue is not all-encompassing performance metric

Key Metric: RevPAR (Revenue per Available Room) Common performance metric in the hotel industry Best when compared across like time or seasonal periods

RevPAR = Rooms Revenue/Rooms Available Rooms Revenue = revenue generated by room sales Rooms Available = # of rooms available for sale

Also: RevPAR = Occupancy % * ADR (Ave Daily Rate)

Questions

Starting Question: How does the oil and gas industry in

Houston impact hotel performance?

Other questions: Are there other factors that affect

hotel performance in Houston? Is this common in other markets?

Hypothesis

We will find a unique relationship between: Overall employment Employment in the oil and gas

industry Oil prices

...and hotel performance RevPAR

Other factors to consider: Per Diem, PPI, Labor Force, Unemployment rate, YoY changes in factors

Our Process

Step 1: Identify

and Collection Relevant

Data Sources

Identifying Data Sources/Ingestion

Question: What factors affect RevPAR in Houston? RevPAR Data (Smith Travel Research Data) Employment & PPI Data (Bureau of Labor Statistics) Gas Prices Data (Energy Information Administration) Government Per Diem Data (Data.gov)

Data Ingestion: Collected and wrangled all relevant data sets in Microsoft Excel

Data Wrangling and Challenges

Data came from 4 different sources

Challenges in Munging and Wrangling: Making sure all dates and columns were consistent Ensure that all calculated metrics had no null values Ensure that all data sets could be read by Python to

produce a full scatter matrix plot

Computation & Analysis

First: Use Python (Pandas and MatPlotLib) to look at all factors together and generate a Scatter Plot Matrix

Then: compute regressions for each factor for Houston. If it rendered a useful result, following up and compare to Chicago, New York, and Denver.

Scatter Plot Matrix

Regression

Dependent Variable: RevPAR

Tableau: Linear and Log regressions (similar results)

Independent Variables: X1 Govt. PerDiem,  X2 Price Per Gallon of Gas, (RLin=0.00; RLog= 1.383e-05)  

X3 PPI for Accommodations Industry, X4 YoY Change in PPI,  X5 YoY Change Mining Employment, (RLin=0.303; RLog=0.324)

X6 YoY Chang in Labor Force,  X7 YoY Change Employment,  X8 YoY Change in Unemployment Rate.  

Houston in Context

Houston market is closest to Denver market for RevPAR vs. YoY Change in Mining and Logging Employment Data Houston R Value =

0.303 Denver R value

=0.288

Chicago Discrepancy

Results

Hypothesis: In the Houston market, there is a direct, positive correlation between RevPAR and overall employment, employment in the oil and gas industry, and oil prices.

Results: In the Houston market, there is a moderate correlation between: RevPAR and change in mining employment

Applications:Ideally we would investigate more market variables as stepping stones to develop a predictive mode.

Macro level: such a predictive model would indicate performance for an overall market given key factors.

Micro level: this type of model would assist operators and owners in pricing strategies to help them outperform their peers.

Questions?