Transcript Science of Hotel Optimization

Science of Hotel Optimization
Rooms Revenue Workshop
Day 1: Data
Day 2: Analysis
Day 3: Optimization
45 minute periods.
15 minute break
every 45 minutes.
http://www.forsmarthotels.com/sohodocs
© Origin World Labs
SOHO Day 3
2
Day 3 Objectives
Hour 1-2
 Capacity Control
Hour 3-4
 Dynamic Pricing
 Micro-Optimization
© Origin World Labs
SOHO Day 3
3
OWL’s vision for The Big RM Reset
Clerical RM
Distribute the Right
Rates and Manage
Inventory.
© Origin World Labs
Analytical RM
To take data, to be able to
understand it, to process it,
to extract value from it, to
visualize it and to
communicate it.
SOHO Day 3
4
Data Science Elements
Data Science
Analytics RM
Tools
Disciplines
Prescriptive
Optimization
Excel, Solver
Economics,
Operations
Research
Predictive
Classification and
Analysis
Excel and SQL
Probability,
Statistics
Descriptive
Data Extraction
and Grouping
MSQuery, SQL
Arithmetic
© Origin World Labs
SOHO Day 3
5
Period Level Dynamic Pricing
© Origin World Labs
SOHO Day 3
6
Capacity Control Optimization - SOHODAY3.xlsx
• Limitations set on the number of units offered to a rate class.
• Prices are provided by the decision maker, not the algorithm.
• Assumes RM has good pricing information.
• Still used in airline and hotel RMS systems.
• Only need to count rooms sold, regardless of rates charged.
© Origin World Labs
SOHO Day 3
7
Standard Deviation
Want to know how spread out the data points are.
Start with the average to measure how far data spreads out.
STDEV.S(data set)
© Origin World Labs
SOHO Day 3
8
Standard Deviation of Rooms Sold by Period and Rate Class
© Origin World Labs
SOHO Day 3
9
Frequency Actual vs. Normal
Normal
Actual
© Origin World Labs
SOHO Day 3
10
Normal Frequency in Excel
Given an average and a standard deviation, you can get the
probability that any # of rooms will be sold using.
1 - NORM.DIST(number of rooms, average, standard deviation, TRUE)
Given an average and a standard deviation, you can get the # of
rooms that will be sold with a certain probability.
NORM.INV(specific probability, average, standard deviation)
© Origin World Labs
SOHO Day 3
11
Expected Value
If the scenario plays out many times.
Reward
x
Chance of Reward =
Rational, Long term Expected Value
(Law of Very Large Numbers)
Core Assumption of all Decision Sciences
The Blue Pill
© Origin World Labs
SOHO Day 3
12
Lottery – Tax on people that don’t know math.
Powerball odds 1/173,000,000 = .000000578% chance of winning.
Costs $2 to play
($150MM - $2) * .000000578% = $.86
- $2 * .9999994% = - $2
Rational Expectation
© Origin World Labs
SOHO Day 3
-$1.14
13
Heuristic – Rule of Thumb
•
•
•
•
•
Easy to calculate and implement.
Used for practical applications.
Based on experience.
Not guaranteed to be optimal.
Common Sense.
© Origin World Labs
SOHO Day 3
14
Capacity Control Pricing Rule
Class 1 is the
highest priced class.
P1 > P 2 > P 3
Switch to higher class
when Expected value is
equal or higher.
© Origin World Labs
SOHO Day 3
15
Capacity Control Algorithms
• EMSRB
• Littlewood’s Rule
• Dynamic Programming
© Origin World Labs
SOHO Day 3
16
Micro-Segmented Dynamic Pricing
SOHODAY3b.xlsx
Period
Room Type
PMS Dimensions
Channel
Company
Rate
Accuracy
© Origin World Labs
SOHO Day 3
17
A Better Demand Curve
Remove Outliers
Low
Limit
Rate
© Origin World Labs
Avg
Gross
Rate
SOHO Day 3
High
Limit
Rate
18
Dynamic Pricing Analytic Tables – SOHODAY3.xlsx
Frequency Tables
Shows the average number
of times a rate was sold per
day per period.
© Origin World Labs
Std Deviation Tables
Allows us to calculate the
upper and lower limit rates
for analyzing the demand
curve.
SOHO Day 3
19
SQL Statistical Functions
Column being analyzed goes inside ()
COUNT(): returns the population (or sample, depending on
the row source)
SUM(): returns the sum of the values in a set
AVG(): returns the mean
STDEV(): returns the standard deviation of a sample
VAR(): returns the variance of a sample
© Origin World Labs
SOHO Day 3
20