Supply%20Chain%20Dynamics

1
Supply Chain Dynamics
N. R. SRINIVASA RAGHAVAN Asst. Professor Departmen
t of Management Studies IISc raghavan_at_mgmt.iisc.er
net.in
2
Introduction

• Oliver and Webber coined SCM in 1982
• Their Thesis
• Top Management alone can ensure non-conflicting
  functional objectives along SCN
• An integrated systems strategy be developed and
  implemented
• Coordination of M-I-F flows is challenging yet
  rewarding

3
Research in SCM

• Research in Integration and Coordination started
  much before 1982
• Channel Research (Alderson 1957)
• Collaboration and Cooperation (Bowersox 1969)
• Location and Control of Inventories in SCN's
  (Hanssmann 1959)
• Hierarchical Production Planning (Hax and Meal
  1975)
• Bullwhip effect in SCN's (Forrestor 1958)

4
1. Channel Research

• Alderson Market Behavior and Executive Action
  (book) 1957 Principles of Postponement
• Product postponement occurs in two ways
• Manufacturing PP Changes in form and identity
  occur _at_ the latest possible point in the SCN
• Logistics PP Changes in inventory location occur
  _at_ the latest possible point in time
• HP Deskjet Printers Lee et al., Interfaces,
  1995.
• Power cords, Voltage requirements, fonts, etc.

5
Channel Research (contd)

• Advantages of PP
• Hedge against uncertain customer demand
• Reduce inventory holding cost
• Reduce Logistics/Warehousing costs
• Minimize imbalance in stock distribution
• Eliminate stages in Manufacturing
• Eg packaging, customer does assembly etc.

6
Features of PP

• Loss of Economies of Scale!
• Requires quick set ups and agile procurement
• Reduced risk of product obsolescence
• Requires increased capability to process,
  transmit, and deliver orders
• Product should be "DFPP"
• (Should be technically and economically feasible)

7
Channel Research CODP

• The location of decoupling point depends on the
  product BOM A, V, X, T
• Relevant more for X and T
• Trade-off between
• Inventory holding costs increase _at_ 20 from L2R
• Estimated based on expected inventories at
  echelons given by the inventory control policies
• Cost of Lost Sales increase _at_ 20 from R2L
• Estimated based on given demand distribution and
  probability of stock outs

8
2. Research on Collaboration and Coordination

• Bowersox Physical Distribution Development, J of
  Marketing, Vol 33, 1969
• Individual objectives of different functional
  units within a firm may counteract overall
  efficiency
• Manufacturing Long production runs
• Procurement Lowest procurement costs
• Marketing FGI Staging and infinite assortments
• Finance Low inventories
• Logistics FTL's

9
Research on Collaboration and Coordination (contd)

• Tom Malone (MIT, CCS) Computers, Networks, and
  the Corporation, Scientific American, 2653, 1991
• Compares performance of various organization
  structures wrt organizational goals and KPI's
• Information Sharing (Transparency) using ICT
• SCM/ERP Solutions
• B2B Markeplaces
• B2C and CRM

10
3. Location and Control of Inventories

• Hanssmann Optimal inventory location and control
  in pdn/distbn networks, Mangement Science, 74,
  1959
• Analytical model (DP) of interacting inventories
  with three serial inventory locations
• Each location follows periodic rev, order up to
  policy
• Lead time is a positive multiple of review period
• Customer demand is Normal
• Trades off shortage costs with inventory holding
  costs considers revenues as a function of
  delivery time

11
Location and Control of Inventories (Contd)

• Supply Rationing Problem Given shortage in
  supplies, how to allocate stock across echelons
• Threshold policies for high priority customers
  (Ha 1997)
• Minimize total imbalance in stock distribution
  s.t. service level constraints (Van der Heijden
  1997)
• Hundreds of articles in various journals
  including OR, MS, EJOR, JORS, IJPR, IJPDLM, JOM,
  etc.
• Logistics given scant treatment in IC problems
  assumed deterministic
• Results in failure of such models in practice

12
4. Hierarchical Production/Distribution Planning

• Hax and Meal Hierarchical integration of
  production planning and scheduling, Studies in
  Management Sc., Vol 1, 1975.
• Provide effective decision support for different
  DM levels within a hierarchical organization
• Based on the following scheme
• Decompose to get hierarchical structure
  (Stgc-Tac-Opn)
• Do Aggregation where possible (eg. Forecasts
  agg. on time,products,markets Capacity agg. On
  resources)
• Hierarchical coordination (by setting
  targetsgetting f/b)

13
Hierarchical Planning (Contd)

• Other elements
• Model building for each decision unit using a
  mathematical model (ILP/DP/QN/PN) keeping
  solvability in mind
• Model solving Solution procedure detailed to get
  the plan
• Cohen and Lee, OR 1989 Solved an integrated
  inventory/production/distribution model
  hierarchically
• Software Agent based frameworks available
• Eg. A-Teams of IBM

14
5. Bullwhip Effect in Supply Chains

• Forrestor Industrial Dynamics, HBR, 364, 1958
• First research paper to illustrate systems
  dynamics in SCN's
• Base for developing Distribution Games
• "BWE" coined by PG
• BWE describes the increasing amplification of
  orders occuring within a SC
• Resembles a whip lash
• Occurs even if end-item demand is fairly stable!
• Forrestor studied a simulation model of the
  simplest tandem supply chain with four entities
  Retailer, DC, W/H, Plant

15
Bullwhip Effect (contd)

• Assumptions of Forrestor's model
• Each entity can make use of only locally
  available info
• Time delays between ordering and receipt of order
• It takes 3 weeks for retailer to process the
  order, half a week to transmit it to DC
• The DC takes 1 week to process the order and one
  week to ship to the retailer, who takes one week
  to ship to end customer assumptions for other
  entities likewise..
• To study impact of a one time 10 change in
  retail sales on orders placed and inventory levels

16
Bullwhip Effect (contd)

• Forrestor's results
• "...A sudden 10 increase in retail sales implies
  a peaking of 34 on orders _at_ 14th week in factory
  w/h, resulting in factory output peaking in 21st
  week (including a 6 week lead time) by a whopping
  45.."
• Amplified and out of phase fluctuations in
  ordering and inventory levels
• Avoidable inventory and shortage costs Unstable
  system
• RELEVANT EVEN TODAY!
• Replace week by day in the above analysis

17
Bullwhip Effect Some Illustrations
18
Bullwhip Effect Some Illustrations
19
Bullwhip Effect (contd)

• Causes of BWE (Empirical Lee et al 1997
  Analytical Chen et al 2000)
• Demand Signal Processing (frequent updates of
  forecasts only next echelon orders considered)
• Order Batching (to realise logistics EoSReducing
  order processing costs)
• Price Fluctuations (resulting in over-reactions)
• Supply Rationing (Proportionate rationing
  unrestricted order acceptancefree return policy)

20
Counter-Measures for BWE

• Avoid multiple demand forecasts
• Order based on ultimate customer demand
• Use EDIPOSVMI
• Choose a good forecasting method (PLC has a major
  say)
• Move from decentralized DM to centralized
  planning (visibilitycontrol is better)
• Remove layers in channel if possible
• Eg HP, Apple, IBM, PG/Walmart

21
Counter-Measures for BWE (contd)

• Break order batches
• Increase frequency of ordering (OP costs reduced
  by EDI)
• Resort to standardization to minmize OP costs
• Use 3PL to make small batch replenishments
  economical
• Aggregate across retail outlets to utilize FTL
  EoS
• Reduce safety stocks by cutting lead times
• Eg 3PL using Fedex, PG

22
Counter-Measures for BWE (contd)

• Stabilize prices
• EDLP (PG)
• Special purchase contracts
• Eliminate shortage gaming
• Allocate based on past sales (Sun)
• Share capacity and information (HP, Motorola)
• Limit flexibility wrt time (HP, Seagate)

23
The Distribution Game

• Sterman Modeling managerial behavior-Misperceptio
  ns of f/b in a dynamic DM expt, Management
  Science, 353, 1989
• "Beer Distribution Game"
• Bounded rationality depicts decision makers
• Orders based on current inventory status, amount
  ordered by direct successors, past performance
• Over-reaction increases steadily towards u/s end
  of SCN