Transcript Document

Decision Support in Travel
Procurement
Applying Cluster Analysis to
Hotel Data
Goal: Select the best hotel chains for a
corporate travel program. Two factors:
• Coverage
• Measuring the potential fit
o Map the demand – known travelers' destinations
o Map the supply – location of hotel properties
• Cost
• Bargaining aggressively yet credibly
o Understand competitive alternatives
o Make credible threats and promises
Problems: Proximity and Comparability
• Traditional proximity problems
o Artificial boundaries - state, county, city, zip code
o Impractical distances
o Poor sorting/ranking/grouping options
• Comparability is easily solved only after proximity is solved
Start with Raw Hotel Data
Mix Together
Add More Ingredients, Then
Mix Into a Very Useful Batter
Slide 1 detail (Arial 44)
Necessary, but not Sufficient
Key Ingredients Include
Tools and the Human Factor
High Quality
Data
Ingredients
1-1/2 cups all-purpose flour
2 teaspoons baking powder
1/4 teaspoon salt
1/2 cup (1 stick) unsalted butter,
softened
3/4 cup sugar
2 large eggs
1/2 cup reduced-fat sour cream
2 tablespoons milk
1/2 teaspoon vanilla extract
5 chocolate and creme sandwich
cookies (such as Oreos), broken up
Frosting:
2 cups confectioners' sugar
1/4 cup (1/2 stick) unsalted butter,
softened
1/4 cup reduced-fat sour cream
1/4 teaspoon vanilla extract
+
Right
Tools
+
Clever
Analytics
Directions
1. Heat oven to 350°. Line 36 indents of
mini muffin pans (using 3 pans total) with
paper or foil liners. If you have only one
or two pans, bake batter in batches.
2. In small bowl, whisk flour, baking
powder and salt. In large bowl, beat
butter until smooth. Beat in sugar until
fluffy. Beat in eggs, one at a time.
3. On low speed, alternately beat flour
mixture and sour cream into butter
mixture, beginning and ending with flour.
Add milk and vanilla; fold in cookie
pieces. Divide batter among prepared
cups; for ease, place batter in a resealable plastic bag, snip off a corner
and pipe into cups.
4. Bake at 350° for 15 minutes, until tops
spring back when lightly pressed.
Remove cupcakes to a rack; let cool.
Slide 1 detail (Arial 44)
From Hotel Mapping…
Slide 1 detail (Arial 44)
To Company-specific Hotel Clusters
Slide 1 detail (Arial 44)
Including No-Stay Hotels
Analytical Approach
• Construct clusters
• Identify and map High-stay hotels
• Build clusters in Matlab, SAS, etc.
• Within each cluster
o Map each cluster's centroid
o Apply a maximum radius limit for each cluster
 Those hotels beyond the radii are "Orphans"
o Bring in the Low-stay and No-stay hotels
• For each hotel within a cluster:
o Allocate "Fair Market Share“ to quantify coverage
o Assign a Buyer Power score (similar to Herfindahl Index)
Calculate Each Chain’s Coverage
Hotel Chain Coverage
Core Markets
34%
24%
17%
16%
Hotel Chain Coverage
5%
All Markets
29%
Marriott
Hyatt
Hilton
Starwood
IHG
24%
21%
15%
9%
Hotel Chain Coverage
Core Markets
Secondary Markets
40%
Marriott
34%
32%
24%
17%
8%
Marriott
16%
5%
Hyatt
13%
5%
Hilton
Starwood
IHG
Hyatt
Hilton
Starwood
IHG
Plot Suppliers on 2 Key Dimensions
Chain Coverage in All Markets
Chain Sourcing Map
30%
Hilton
Starwood
Marriott
25%
20%
Hyatt
15%
IHG
10%
5%
0%
-
2.0
4.0
6.0
Buyer Power (10 = Very High)
8.0
10.0
Derive Expected Program Value Scores
Chain Sourcing Map
Core Markets
34%
24%
17%
16%
5%
Marriott
Hyatt
Hilton
Starwood
IHG
Expected Program Value
3.5
2.1
Marriott
Hyatt
1.5
Hilton
Starwood
Starwood
Hilton
30%
Marriott
25%
20%
15%
10%
IHG
Hyatt
5%
0%
-
2.0
4.0
6.0
8.0
Buyer Power (10 = Very High)
Likely Best Chains:
•Marriott, Starwood
•Marriott, Starwood, Hilton
4.5
1.5
Chain Coverage in All Markets
Hotel Chain Coverage
IHG
10.0
High Quality
Ingredients
Clever
Chefs and
Recipes
Proper Tools
Thank you!
Contact
Dan Pirnat, VP & GM
TRX Travel Analytics
(O) +1 440 236 3561
(e) [email protected]