RM Coordination and Bid Price Sharing in Airline Alliances: PODS Simulation Results

1
RM Coordination and Bid Price Sharing in Airline
AlliancesPODS Simulation Results

• Peter Belobaba
• Jeremy Darot
• Massachusetts Institute of Technology
• AGIFORS YM Study Group
• Bangkok, May 9-11, 2001

2
Outline

• PODS Alliance Network Characteristics
• Airline A vs. Airline BC Alliance
• Revenue Impacts Without RM Coordination
• Unequal Revenue Benefits to Alliance Partners
• Differential Revenue Gains for DAVN vs.
  Probabilistic Bid Price
• Valuation of Code Share Passengers in RM Models
• Local Fare, Total Fare or Y-Fare Prorated Inputs
• Bid Price Sharing Between Alliance Partners
• Revenue Benefits of Dynamic Information Exchange

3
PODS Network Geography
1
H1(41)
2
21
3
4
5
25
6
24
23
27
26
7
31
28
30
8
32
29
33
22
9
11
35
34
38
10
12
14
15
13
H2(42)
16
36
17
18
37
19
39
20
40
4
A vs. B/C Alliance in PODS

• Airline A remains unchanged, and still operates
  from the MSP (northern) hub
• Complete service from 20 west cities to 20 east
  cities
• Former Airline B is split into two unequal
  alliance partners, which both operate from the
  DFW (southern) hub
• Airline B operates longer haul flights from/to 10
  northern cities on each side of hub, and also
  operates interhub flights.
• Airline C operates shorter haul flights from/to
  10 southern cities on each side of hub.
• Code-share flights provided for all BC
  interline connections

5
Airline A
Airline B
Airline C
6
Alliance Network Characteristics

• Asymmetric alliance partners
• Airline B has longer-haul flights to southern
  hub, lower load factors, but carries more
  passengers and RPMs
• Airline C has shorter-haul flights and higher
  average load factors
• Implications for code-share paths
• Airline B benefits from increased code-share
  traffic, given lower load factor and greater
  proportion of distance flown (and revenue
  allocated) on own operated aircraft
• For Airline C, code-share paths have higher
  alliance revenue value than own connections, but
  load factors are higher and revenue allocation is
  smaller (based on Y-fare proration revenue
  sharing)
• Valuation of code-share paths is critical to RM
  performance

7
Base Case EMSRb Fare Class YM
2 airlines A vs. B
A vs. B/C Alliance
8
Simulation Parameters

• RM methods
• Airline A uses EMSRb Fare Class Yield Management
  (FCYM)
• Airline B and C using
• BASE CASE EMSRb Fare Class Yield Management
  (FCYM)
• Displacement Adjusted Virtual Nesting (DAVN)
• Probabilistic Network Bid Price Control (ProBP)
• RM model input fares for code share paths in
  connecting markets served by Airlines BC
• Local fare valuation (use local fare of same fare
  class)
• Alliance revenue sharing based on Y-fare
  proration of fares

9
Use of DAVN by Alliance Partner(s)

• Standard DAVN in PODS defined as follows
• 8 leg-specific virtual buckets based on total
  fare minus network displacement on own legs
  applies only to own-connect paths
• No displacement for code-share paths, which are
  treated as local paths, and valued at local
  fare value of same fare class
• Implications for code-share revenue management
• Own-connects can be nested higher or lower than
  locals/code-shares, depending on down-line
  displacement
• Code-share paths are subject to same virtual
  bucket EMSRb booking limits as local paths

10
Alliance DAVN vs. FCYM Base Case
RM Method Used by B/C
11
Change in Traffic Mix DAVN/FCYM(compared to
Base Case)

• B gains (0.78) revenue, C gains (0.20),
  Alliance gains (0.52)
• DAVN allows B to increase own-connects, while
  locals and code-shares both decrease
• Use of displacement costs means B takes good
  connects
• Leg-specific virtual buckets means less
  preference to locals and code-shares on long legs
• Reduced code-share flow in turn benefits C

12
Change in Traffic Mix FCYM/DAVN

• B gains (0.30) revenue, C gains (0.48),
  Alliance gains (0.38)
• DAVN reduces own-connects for C, due to higher
  ALF
• Mostly low class spill
• B still gains revenue from increased
  code-shares, given lower ALF and favorable
  revenue sharing

13
Results Alliance DAVN vs. FCYM

• Alliance partners benefit from one or both using
  DAVN
• Revenue gains even for airline with FCYM when one
  partner moves to DAVN OD control
• Better control of connecting traffic by DAVN
  partner can lead to increased revenue for FCYM
  partner (from either more or fewer code-share
  passengers)
• Revenue gains are greatest for both partners and
  the alliance when both use DAVN.
• Because of its higher market share and lower load
  factors, B benefits more from switching to DAVN.

14
Use of ProBP by Alliance Partner(s)

• Standard ProBP in PODS defined as follows
• Nested Probabilistic Convergence Algorithm
  (Bratu, 1998)
• Iterative proration of all ODFs on each leg in
  network to find convergent probabilistic leg bid
  prices
• Additive bid price control for all local and
  connecting paths
• Implications for code-share revenue management
• Own-connects subject to sum of bid price values
  over both legs operated by single partner
• Code-share paths are valued at local fare levels
  for optimization, and controlled only by bid
  price of own (operated) leg

15
Alliance ProBP vs. FCYM Base Case
RM Method Used by B/C
16
Alliance ProBP vs. FCYM

• ProBP does not perform as well as DAVN for the
  alliance carriers using separate RM systems
• Major difference is performance of PROBP for
  Airline C
• Revenue gains, in percent over FCYM baseline
  case

17
Change in Traffic Mix PROBP/FCYM (compared to
Base Case)

• B gains (0.70) revenue, C loses (-0.21),
  Alliance gains (0.29)
• B carries fewer own-connects than with DAVN
  (slide 11)
• Suggests PROBP bid prices are affected by
  code-shares, now treated as local paths
• Code-shares and locals both decrease
• Cs losses come from change in mix, as Q locals
  and own-connects replace previous B and M pax

18
Change in Traffic Mix FCYM/PROBP

• B gains (0.12) revenue, C gains (0.09),
  Alliance gains (0.11)
• Use of PROBP by Airline C leads to large
  decrease in own-connects
• Loss in all classes except Y
• Distortion of PROBP bid prices clearly more
  substantial
• ALF is notably lower than under DAVN
• Small gains for B from increased code-shares
• Spread across all classes

19
Summary No RM Coordination

• Differences in O-D benefits to alliance
  partners
• Weaker airline can actually see its revenues
  decrease when larger partner moves to O-D control
• Unequal impacts even when both partners use same
  O-D method
• Depends on network characteristics and valuation
  of code-share passengers in RM optimization
• Differences between DAVN and ProBP revenue
  gains
• DAVN shows robust revenue gains in alliance
  combinations
• In contrast, PROBP falls short in revenue
  performance
• Bid price values are being distorted by treatment
  of code-share passengers as locals impact is
  greater on probabilistic bid prices

20
Valuation of Code Share Passengers

• Previous results assume connecting BC
  code-share passengers valued at local fare for
  separate RM optimization.
• We also compared alternative approaches to
  valuation of code-share passengers in connecting
  (code-share) markets served by Airlines BC
• Local fare discount (use local fare of same fare
  class)
• Total path fare (no discount)
• Y-fare proration of connecting paths (same as
  revenue-sharing agreement assumed in all PODS
  results)

21
Code Share Valuation DAVN/DAVN
22
Summary Code Share Valuation

• Valuation of code share passengers for RM
  optimization and control has a significant effect
  on DAVN revenue gains
• Using total fares leads to the highest revenue
  gains for Airline B and total alliance, as it
  increases code-share traffic
• Using local fares leads to slightly lower, but
  more evenly shared revenue benefits
• Using Y-ratio fare inputs leads to the smallest
  alliance gains, but favors airline C
• At higher demand factors, relative revenue gains
  change
• In this network, using Y ratio fare inputs leads
  to higher revenue gains than either total fares
  or local fares as RM inputs

23
Information Sharing between Partners

• Each alliance partner still performs RM
  optimization for own network separately
• Calculate leg displacement costs if using DAVN
  O-D controls
• Calculate leg bid prices if using ProBP O-D
  controls
• Separate network optimization assuming either
  local fare or total fare valuation of code-share
  connecting passengers
• Bid price sharing involves exchange of leg bid
  prices or displacement costs between alliance
  partners
• Modeled in PODS as dynamic sharing of values for
  each leg in network, by booking period with a one
  period lag

24
Alliance Information Sharing in PODS
25
Alliance DAVN with Info Sharing

• When both alliance partners use DAVN,
  displacement cost sharing improves the results of
  DAVN compared to FCYM.
• Network revenue value for code-share paths is
  total fare minus displacement costs on own and
  partners connecting legs
• Both Airlines B and C achieve greatest revenue
  gains over base FCYM, as does the total alliance
• Revenue gains, in percent over baseline case

26
Displacement Cost Sharing DAVN/DAVN
27
Alliance ProBP with Info Sharing

• Bid price sharing improves the results of
  alliance ProBP dramatically compared to FCYM
  base
• With bid price sharing, ProBP performs better
  than DAVN (in contrast to previous results
  without bid price sharing).
• The gains are higher when total fares are used as
  inputs to RM optimization model.
• Revenue gains, in percent over baseline case

28
Bid Price Sharing ProBP/ProBP
29
Alliance with DAVN/ProBP

• Even if partners use different RM methods and
  separate optimization, benefits of information
  sharing are substantial
• Benefits of RM coordination exceed differences
  between network optimization methods
• Revenue gains, in percent over baseline case

30
Bid Price Sharing DAVN/ProBP
RM Method Used by B/C
31
Summary of Findings

• Revenue gains of O-D control can be affected by
  alliances
• With separate and uncoordinated RM, one partner
  can benefit more than the other, even causing
  other partners revenues to decrease
• O-D methods perform differently (DAVN vs. ProBP),
  depending on network structure and valuation of
  code-share passengers
• Valuation of code share passengers for RM
  optimization affects both total revenue gains and
  partner revenue shares.
• Information sharing significantly improves the
  performance of OD control, even if partners use
  different OD methods.
• In most cases, sharing of bid prices or
  displacement costs reduces the discrepancy
  between the revenue results of the partners.
• Currently limited by technical and possibly legal
  constraints.

32
PODS Future Work Alliance RM

• Impacts of uncoordinated RM in alliances
• Additional RM methods (e.g., Heuristic EMSR Bid
  Price)
• Code-share passenger valuation for optimization
  vs. control
• Valuation of own code-share passengers vs.
  partners
• Information exchange between alliance partners
• Less frequent exchange of bid prices/displacement
  costs
• Bid price inference by one airline of partners
  bid prices from CRS leg/class availability
• Larger, less symmetrical alliance network
• Longer-haul international flights smaller
  code-share partners