Innovation distinguishes a leader from a follower. Carlson Rezidor Hotel Group stands out as an innovator that’s setting new industry standards with its next-generation pricing and revenue management approach. Accordingly, Carlson was named as a finalist for the 2012 Franz Edelman Award for its exemplary achievement in leveraging both operations research and analytics to improve the real performance of its business. Carlson partnered with JDA Software to create Stay Night Automated Pricing (SNAP), a next-generation revenue management approach that maximizes revenue by recommending optimal prices based on real-time competitor rates and customers’ willingness to pay. The company’s one-two punch resulted in a consistent revenue premium of 2-4 percent for U.S. properties.
2. in partnership with:
Carlson Rezidor Hotel Group maximizes revenue through
improved demand management and price optimization
Franz Edelman Award for Achievement in Operations
Research and the Management Sciences
4. 9 th
Largest Hotel Company in the World
Lead position in Europe, India and Russia
53
341
640
#
1
#
1
36 32
41 98
52
1
#
26
2011 Total = 1,319
5. Six Brands Span The Entire Service Level Range
Luxury room distribution by brand
Upper 23% 0%
Upscale 34%
Upscale 15%
5% 23%
Upper
Midscale
Midscale Missoni Park Plaza
Radisson BLU Park Inn
Economy Radisson Country
Inns & Suites
6. sm
Country Inns & Suites By Carlson
Luxury
Upper
Upscale
Upscale
Upper
Midscale
Midscale
7. Park Inn by Radisson
Luxury
Upper
Upscale
Upscale
Upper
Midscale
Midscale
15. Hotels Are Rarely Full
U.S. average occupancy (STR)
64.0
63.0 63.1 62.8
62.0
60.0 60%
59.8 59.9 60.0
58.0
57.6
56.0
54.0 54.6
52.0
50.0
2005 2006 2007 2008 2009 2010 2011E 2012E
16. A Challenging Time To Price Correctly
• Elementary pricing approach
- Decentralized
- No forecasts
- No optimization
• Revenue-dilutive pricing behaviors
- Chasing competition
- Locking on lowest price
- Flat rates
17. Solution Approach and OR Methodology
Pelin Pekgün, Ph.D.
lead scientist and manager, analytical services, JDA software group
18. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
19. Solution Overview
Market Response Competitive
Demand Forecasting
Model Modeling
• Demand Patterns • Customer Segmentation • Competitor prices and
• Seasonality and Booking Pace • Price Elasticity Modeling availability
• Special Events and External • Promotional Effects • Market Reference Price
Factors
Constraints Price
• Capacity
• Network effects Optimization
• Business Rules
Price Recommendations
20. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
21. Time Series Demand Forecasting
• Forecasting Horizon: 120 days
• Demand Forecasting Unit (DFU): Arrival
- Hotel Dates
- Rate Segment
- Length of Stay (LOS) 0
- Day of Week
Days Left
- Booking Interval
• Methodology: Multiple Linear Regression
- Seasonality (Fourier series)
- Special Events (Concerts, Sporting events, City convention, etc.) 120
- Holidays (New Year’s Day, Christmas, Thanksgiving, etc.
22. Hierarchical Forecasting and Reconciliation
Aggregate level:
Parent DFU Hotel, Day of Week, Booking Interval
Low level:
Child (Opt.) Child (Opt.) Hotel, Day of Week, Booking Interval,
DFU DFU Rate Segment, LOS
• Higher number of terms for Parent DFU to capture seasonality
• Lower number of terms for Child DFUs to address data sparsity
• Top down reconciliation:
23. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
24. Hurdle Rates vs. Optimal Rates
“What price to charge to
generate the most return on
remaining inventory?”
25. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
26. Why Stay Night Prices?
• Variety of prices to optimize difficult to review by the users:
Rate Segments Arrival Dates Booking Intervals LOS Patterns
Rate Menu Menu Stay Night Rate
Segment Type Offset
4/14 $250
Base Rate Mult. 1.0
4/15 $220
Value Added Add. +$20 4/16 $180
10% Off Mult. 0.9 Arrival Date: 4/14, LOS=3;
Rate = 250+220+180 = $650
Rates to Optimize
20% Off Mult. 0.8
27. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
28. Price Optimization Formulation
DFU price as a function Penalty to
of the stay night price Unit Ancillary minimize price
via rate menu Cost Revenue fluctuations
Maximize Revenue minus
Penalties
State Trajectory of Bookings
with Survival
Price Sensitive Demand
(Bookings-to-come) Function
Capacity
Constraint
Business Rules to Bound
Price Changes
Fluctuations from
the current price
…with additional monotonicity and
non-negativity constraints, etc.
30. Price Optimization Structure
• Network optimization • Floating-point price recommendations
• Linear price demand curve • Fixed rate segments controlled for
availability
• Quadratic programming
• Group forecast pre-deducted from
• CPLEX solver room availability
• Supported by a discrete-time
step simulator – Constrained
Forecast Evaluation
31. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
32. Market Reference Price (MRP) Computation
Prepare Competitor Price
Shops
Construct MRP applying
user weights
Use MRP directly in price
optimization
Baseline Own Price Market
Forecast Elasticity Reference Price
33. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
34. Data-Driven Competitor Set Identification
Step I. Attribute-based classification (Score-based Benchmarking)
- Location
- Size
- Brand
- Service
- Amenities
MAPE: 18.73% 4.61%
35. Data-Driven Competitor Set Identification
Step II. Constrained regression (Automated Weight Estimation)
– Dynamic price shops analyzed over time
– Competitor weights constrained to add up to 1
User Computed
Competitor
Weights Weights
A 0.2 0.27
B 0.2 0
C 0.2 0.46
D 0.2 0.27
E 0.2 0
36. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
37. Statistical Price Elasticity Estimation
• Regression based methods
• Requires historical data
- Demand history
- Actual prices paid by customers
- Competitive price information
• Does not help with
- Property initialization
- Non-dynamic markets
- New properties
38. Business-Driven Elasticity Estimation
Distributes elasticity values across rate segments given
• an ordering of segments based on expected price sensitivity
• upper and lower elasticity bounds (e.g., -0.5 to -3.5)
• a centering value (e.g., -1)
• expected remaining demand forecasts
RATE SEGMENT Median Rate Forecast Price Elasticity Weighted Forecast
Value Added $145.00 100.22 -0.50 -50.11
Base Rate $127.00 944.97 -0.66 -619.55
Special Rate $122.55 102.00 -0.81 -82.75
10% Off $113.40 838.26 -0.97 -810.52
20% Off $101.60 1427.16 -1.12 -1602.04
25% Off $95.25 711.25 -1.28 -909.10
50% Off $49.50 59.72 -1.83 -109.51
Totals 4183.58 -4183.58
Total Weighted Forecast / Total Forecast = -1.00 (Centering Value)
39. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
41. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
42. …Must Get Results by 8am
Automated Processing User Interactions
Forecast
Optimize Prices Manage System
Demand
Parameters
Shop Process
Competitive Inventory Data
Rates
Compute Process
Market Booking Data
Reference
Prices Maintain
Competitors Review Rate
Update Recommendations
Elasticities Upload Prices
43. Processing by Zone
6PM
Users en route to work
System optimizing prices
Users working recs
Regular rate uploads
Users sleeping
System idle
44. Processing by Zone
2AM
Users working recs
Regular rate uploads
Users sleeping
System optimizing prices
Users en route to work
System is idle
45. Processing by Zone
8AM
Users sleeping
System forecasting demand
Users en route to work
System forecasting demand
Users working recs
Regular rate uploads
System forecasting demand
46. Processing by Zone
6PM
Users en route to work
System optimizing prices
Users working recs
Regular rate uploads
Users sleeping
System idle
47. Enterprise Architecture
Internet and
Users Application Database
Load Balancer Servers
Servers
and Web Tier
Grid Servers
48. Key Solution Components
JDA Demand JDA Travel Price Optimization (TPO)
Internet and
Users Application Database
Load Balancer Servers
Servers
and Web Tier
Grid Servers
49. TPO is an extensible tool that has been applied in
several industries
Carlson Rezidor, 2010 Passenger rail, 2011 Yacht rental, 2011 Golf , 2012
50. TPO is an extensible tool that has been applied in
several industries
Group
Evaluation
51. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
52. Driving Revenue Upwards – An Ongoing Journey
Product and System Rollout - Worldwide
Demand Forecasting Development
2006 2007 2008 2009 2010 2011 2012
Price Optimization Prototyping Rollout - Americas Group Evaluation
and Market Trial
53. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
54. Benefits and User Acceptance
Kathleen Mallery
director, revenue optimization, planning and development
55. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
59. Proved the Value – Simple Yet Effective
Series 1
+5%105
Hotels using the +4%
Incremental Year-over-Year
104
prototype realized a +3%103
Incremental
2-4% incremental
Revenue
+2%102 year-over-year
Value from tool
revenue improvement +1%101
Baseline
100
99
98
Control Hotels SNAP Pilot Hotels
64. Manage Exceptions
• Look only at changes
• Drill down to determine reason
Previous Recommended Rooms to be Sold at
Rate Rate Recommended Rate
Rooms to be Sold at
Previous Rate
66. SNAP Now Has Many Supporters
• 64% of hotels use SNAP
• 183 hotels auto-pilot
• 35% observed maximum
unit revenue growth
67. User Testimonial
“With SNAP we are now able to see
our future rate mix and see what
[rates] will be displaced. This
allows for interjections before day-
of-arrival, to assure we maximize
every potential source of revenue.”
Randy Smith
Country Inn and Suites
Dothan, Alabama
68. Solution Requirements
Models seasonality and special events
Works on high- and low demand nights
Recommends stay night prices
Minimizes day-to-day price fluctuations
Keeps rates in line with the marketplace
Identifies true competitors
Understands price elasticity
Makes results available by 8 am
Not a custom solution for Carlson Rezidor only
Progressive return on investment
Gains the stakeholders’ trust
Revenue Management in 30 minutes a day
69. Thank You
“Through the first 39 days of 2012, ADR is 38% higher year -over-year while the
average hotel in my market is up 6% - 10%. I attribute the majority of this
increase to SNAP and its autopilot functionality.”
– Himansu Patel, Owner of the Radisson Hotel in Akron/Fairlawn, Ohio
“[SNAP] is easier than other brands have, we love it and have seen a GREAT lift
in ADR due to SNAP, up 6.97 USD”
– Andrew Behnke, GM & Owners Rep for Radisson Buena Park, California (manages multiple hotels brands)