ESD Faculty Lunch Research Talk Mustafa agri Grbz

1
ESD Faculty Lunch Research Talk Mustafa Çagri
Gürbüz

• April 14, 2009
• CTL _at_ MIT

2
Introduction

• BS in Industrial Engineering, Bilkent University,
  Ankara Turkey, 1999
• MS in Industrial Engineering, Bilkent University,
  Ankara Turkey, 2001
• PhD in Operations Management, Michael G. Foster
  School of Business, University of Washington,
  2006
• Faculty member at the Zaragoza Logistics Center
  since 09/2006
• Visiting faculty at CTL until 09/2009

3
Agenda

• Academic Research
• Inventory/Transportation Management in
  Distribution Systems
• Coordinated Replenishment Strategies in
  Inventory/Distribution Systems, with K.
  Moinzadeh and Y. Zhou, Management Science, Vol.
  53 (2), 2007, 293-307.
• Inventory Management under random supply
• Supplier Diversification Under Binomial Yield,
  with M. Fadiloglu and E. Berk, Operations
  Research Letters, Vol. 36 (5), 2008, 539-542.
• Contracting retailer/manufacturer efforts in a
  newsboy setting
• Impact of random deal offerings for perishable
  products under continuous review
• The impact of accountability on the bullwhip
  factor
• Projects at ZLC
• Revenue Management for the passenger rail
  industry
• Measuring carbon footprint due to transportation
  for the European distribution of Print Green
• Investigating Spains potential in distributing
  goods in Europe

4
Coordinated Replenishment Across Retailers
Suppliers

• Mustafa Cagri Gürbüz
• MIT-Zaragoza International Logistics Program,
  Zaragoza, Spain
• Co-authors Kamran Moinzadeh, Yong-Pin Zhou
• University of Washington, Michael G. Foster
  School of Business

5
Distribution Costs!

• Distribution costs are cited as 10 of GDP for
  developed countries, and 20 or more for
  developing countries (a World Bank research paper
  by Bagai and Wilson, 2006)
• Distribution costs represent on average 15 of
  the selling price (Van Damme 2000) in European
  companies
• 32 transportation costs
• 31 inventory costs
• 28 facility costs
• Industry Week Value Chain Survey conducted in
  2005 (www.industryweek.com) ?
• The percentage of respondents stating more than
  10 increase in distribution costs of sales has
  more than doubled since 2003

Borrowed from Dr. Emre Berk
6
Consolidation/Coordination

• Majority of the companies use some form of
  shipment consolidation meaning
• Combining multiple shipments into a single group
  (across time, locations, products) to achieve
  lower costs
• Time based consolidation
• Quantity based consolidation
• Time and quantity based

7
Coordinated replenishment (two-items)
Order trigger at all retailers, combined
8
Challenges

• Use of information to decide
• How to coordinate shipments?
• When to order?
• How much to order?
• GOAL To minimize the overall cost, which is the
  sum of
• Fixed Ordering/Setup,
• Holding/Backorder,
• Transportation.
• The optimal solution to this problem?

9
Analysis

Cost Rate
Inventory level distribution
Inbound quantity distribution
Inbound penalty cost
Holding/Shortage cost
Outbound penalty cost
Ordering cost
Inventory position distribution
Expected cycle time
Outbound quantity distribution
10
Coordination across retailers alone

• Each item is ordered independently
• but retailers are replenished simultaneously
• Policy MII0 The warehouse orders to raise all
  the retailers inventory position to Sj for item
  j whenever
• any retailers inventory position for item j
  drops to sj
• OR
• the total demand at all the retailers for item j
  reaches Qj (for j1,M).

11
Coordination across retailers items (suppliers)

• Policy MISO-1
• Consider Sub-policy j for all j1,2,..,M
• Monitor IP for item j only,
• Trigger Mechanism Replenishment happens
  whenever
• any retailers inventory position for item j
  drops to sj or
• the total demand at all the retailers for item j
  reaches Qj.
• Dispatch Mechanism
• Raise all the retailers inventory position to Si
  for item i when the replenishment is triggered,
• Ask the supplier to ship item i exactly l1i
  (L01-L0i) time units after replenishment is
  triggered (assume L01L0i for all i).
• Evaluate the cost rate for Sub-policy j
• Pick the sub-policy with the minimum cost rate.

12
Illustration of Policy MISO-1
Item 2 is shipped out from Supplier 2
0
t
t?o2
tL01
0
t
tl12
1) Trigger for Item 1 (or Item 2) 2) Raise the
inventory position for Item 1 and Item 2 3) Item
1 is shipped out from Supplier 1
1) Both items arrive at the warehouse at the same
time 2) They are shipped to the retailers
13
Coordination across retailers items (suppliers)

• Policy MISO-2
• Consider Sub-policy j for all j1,2,..,M
• Monitor IP for item j only,
• Trigger Mechanism Replenishment happens
  whenever
• any retailers inventory position for item j
  drops to sj or
• the total demand at all the retailers for item j
  reaches Qj.
• Dispatch Mechanism
• Raise all the retailers inventory position to Si
  for item i and ask the supplier to ship item i
  exactly l1i (L01-L0i) time units for all i after
  replenishment is triggered (assume L01L0i for
  all i).
• Evaluate the cost rate for Sub-policy j
• Pick the sub-policy with the minimum cost rate.

14
Illustration of Policy MISO-2
1) Raise the inventory position for Item 2 2)
Item 2 is shipped out from Supplier 2
0
t
t?o2
tL01
0
t
tl12
1) Trigger for Item 1 (or Item 2) 2) Raise the
inventory position for Item 1 3) Item 1 is
shipped out from Supplier 1
1) Both items arrive at the warehouse at the same
time 2) They are shipped to the retailers
15
Coordination across retailers items (suppliers)

• Policy MISO-3
• Monitor IP for all items,
• Trigger Mechanism Replenishment happens
  whenever
• any retailers inventory position for any item j
  drops to sj or
• the total demand at all the retailers for any
  item j reaches Qj.
• Dispatch Mechanism
• Raise all the retailers inventory position to Sj
  for item j (all items j1,2,..,M) when the
  replenishment is triggered,
• Ask the supplier to ship item j exactly l1j
  (L01-L0j) time units after replenishment has been
  triggered (assume L01L0j for all j).

16
Illustration of Policy MISO-3
1) Item 2 is shipped out from Supplier 2
0
t
t?o2
tL01
0
t
tl12
1) Trigger for Item 1 OR 2 2) Raise the inventory
position for items 1 and 2 3) Item 1 is shipped
out Supplier 1
1) Both items arrive at the warehouse at the same
time 2) They are shipped to the retailers
17
Coordination across retailers items

• Policy MISO-4
• Monitor IP for all items,
• Trigger Mechanism Replenishment happens
  whenever
• any retailers inventory position for any item j
  drops to sj or
• the total demand at all the retailers for any
  item j reaches Qj.
• Dispatch Mechanism
• Raise all the retailers inventory position to Sj
  for item j exactly l1j (L01-L0j) time units after
  replenishment has been triggered,
• Ask the supplier to ship item j exactly l1j
  (L01-L0j) time units after replenishment has been
  triggered (assume L01L0j for all j).
• Assume no trigger will happen for item j for the
  next ljM time units after the inventory position
  is raised to Sj for j1,..,M-1.

18
Illustration of Policy MISO-4
1) Raise the inventory position for Item 2 2)
Item 2 is shipped out from Supplier 2
0
t
t?o2
tL01
0
t
tl12
1) Trigger for Item 1 OR 2 2) Raise the inventory
position for Item 1 3) Item 1 is shipped out
Supplier 1
1) Both items arrive at the warehouse at the same
time 2) They are shipped to the retailers
19
Summary of Coordinated (across items) continuous
review policies
MISO-2 1. Monitor IP for one item only 2.
Internal delay
MISO-1 1. Monitor IP for one item only 2.
External delay
MISO-3 1. Monitor IP for all items 2. External
delay
MISO-4 1. Monitor IP for all items 2. Internal
delay
20
Numerical Results

• No significant difference between coordination
  through internal or external delay
• Policies MISO_3 and MISO_4 perform very closely
• Policies MISO_1 and MISO_2 are good heuristics
• Their performance are pretty close to that of
  Policies MISO_3 and MISO_4 in many cases
• Easier to analyze and compute
• Rankings (best-worst) of the policies are as
  follows (the improvement over the MIIO is given
  in parentheses)
• MISO_3 (2.12), MISO_4 (1.82), MISO_2 (0.59),
  and MISO_1 (0.42)
• Monitoring inventory positions for both items
  help Policies MISO_3 and MISO_4 for higher ?.

21
Numerical Results

• Benefits from coordination across items increase
  for
• More retailers (higher N)
• Larger fixed inbound/outbound ordering costs
  (higher K0 and K/ K0)
• Larger outbound truck capacities (higher C)
• Smaller unit outbound transportation penalty cost
  inbound (smaller ?)
• Smaller difference in transit times from supplier
  to warehouse (larger L02/ L01)

22
Revenue Management?

• What is it?
• Pricing train seats for specific market segments
• Protecting seats for each segment based on demand
  (capacity allocation)
• Why should passenger rail companies use it?
• Unfilled train seats Lost Revenue
• Full trains Lost Revenue
• Why should YOU care?
• Understanding it can help you save money

Borrowed from S. Joiner
23
A Simple Example
No Revenue Management

• Pricing Scheme

Discounting
Revenue Management
Departure Day
Departure Day
Departure Day
-14 Days
-14 Days
-14 Days
-7 Days
-7 Days
-7 Days
50
50
50
30
30
30
20
30
50
(Max 4)
(Max 14)

• Seat Utilization 93
• Total Revenue 1590
• Revenue per Seat 28,90

• Seat Utilization 100
• Total Revenue 2470
• Revenue per Seat 45,33

• Seat Utilization 65
• Total Revenue 1850
• Revenue per Seat 32,50

• Total Revenue Summary
• No RM 1850

Discount 1590
24
Why is this so difficult?

• Data Limitations
• Limited historical data is available
• Historical data does not help understand how
  customers will respond to price changes
• The Rail Network
• Unlike in the previous example, passengers can
  enter and exit the train at various locations
  during a trip
• A seat protected for the Zaragoza-Barcelona leg
  means one less seat is available for the
  Madrid-Barcelona leg

Barcelona
Zaragoza
Madrid
Borrowed from S. Joiner
25
The Research

• Question
• What general guidelines can be established for
  applying revenue management in the passenger rail
  industry?
• Approach
• Using historical passenger data and customer
  surveys from RENFE to understand and predict
  consumer behavior
• Simulation
• Developed a simulation model to see how different
  seat protection and pricing schemes affect
  revenue

Borrowed from S. Joiner
26
Contact information

• Email
• mgurbuz_at_zlc.edu.es
• mgurbuz_at_mit.edu
• Address
• Avda. Gomez Laguna, 25, Planta 1,
• 50009 Zaragoza, Spain
• Phone
• 34 619 44 62 66