Outline of the talk by jlesterback



          “Inventory Control, ERP, and SCM”
                      DORS Seminar
                       April 5, 2001
                    Aarhus University
            Department of Mathematical Sciences

                    Anders Thorstenson

                   Outline of the talk

          Ø General observations on decision-support
            systems for Supply Chain Management

          Ø Inventory Control in SCM

          Ø Examples of Inventory Control models
            for SCM

                 SCM: The Distribution Game
                                 (Jackson, 1995)
      Supplier                                                                       Retailers


              Even for this very simple supply chain the best
           (’optimal’) inventory policy is not known in general!

            Extended Supply-chain Network

                   Supplier                  Supplier                   Supplier

                              Supplier                     Supplier


                                    Central warehouse

                      Distribution center               Distribution center

                      Retailer    Retailer     Retailer      Retailer     Retailer


            Information and Control Systems

     Ø ERP (Enterprise Resource Planning)
       Basic enterprise information and planning modules related to
       a common data base
         (MRP: Material Requirements/Resource Planning)
         (DRP: Distribution Requirements Planning)
     Ø APS (Advanced Planning and Scheduling;
             APSII, APO, ERO, etc.)
       Control modules for active decision support (ERP ’Bolt-ons’)
     Ø SCP (Supply Chain Planning)
       Modules for multi-echelon information and control
     Ø SCE (Supply Chain Execution)
       Operational control and feedback modules
     Ø LES, MES, etc.

          Syndromes, Paradoxes, and Myths
          Ø The 3 OPT syndromes
             § The term optimize seems to have been ‘debased to mean
               try to improve’ (Geoffrion & Krishnan, 2000)
             § Rapid developments make state-of-the-art surveys of the
               optimization software illusive and practically impossible
             § Proprietary software and business lingo make methods and
               algorithms less amenable for evaluation (Cf. the OPT disc.)
          Ø The Decision Support Paradox
             § The more decision support that is provided by the systems
               the higher the competence that is required by the user
               (However, the # of users required might be less)
          Ø The Black Box Myth
             § A complete understanding of methods and algorithms used
               is required by the user (?)

                         An observation:

          “First, the word ’Optimization’ is no longer a bad word
            that cannot be spoken in business, and that means it
            will eventually be used in respectable business
            schools again, as students demand it.”

                                                   M.S. Sodhi,
                                                   OR/MS Today
                                                   October 1999

                       End user views (?)

      “90% of major system implementations are disappointments,
        according to new survey*”
      “75% say decision-support is lacking”
      “The tools, the approaches, and the methods available in the
        systems are not, at present, adequate enough to satisfy the
        demands for harvesting from the heaps of data

                     (Translated from Børsen, Feb. 23, 2000;
                     *) Survey by the PA Consulting Group of 65 ERP
                        implementations in Europe, incl. Scandinavia.)

          Decision support applications in SCM

              Drivers and enablers:
              Ø (Global) Competitive forces
              Ø Supply chain complexities
              Ø Data availability (ERP)
              Ø ICT/IS developments
                 – including the Internet/WWW
              Ø Modeling and solution capabilities (APS)
              Ø Educational level and management

                Decision support for
              ERP, E-business, SCM, etc.

               Decision support systems
               Analytical IT-systems
                 require a decision data base
                         •   management accounting principles
                         •   management policy information
                         •   decision modeling specifications
                         •   aggregation & disaggregation facilities
                         •   external & future data sets
                         •   feedback & GUI features
               (in addition to the requirements for
                  transactional systems; Shapiro,1999)

                        Inventory Control

          Structure and Coordination in SCM
          Ø Structural decisions (long-term, strategic issues), e.g.,
               –    Number and location of facilities and sources
               –    Type and capacity of facilities and sources
               –    Allocation of products to facilities and sources
               –    Transportation systems and modes
               =>   Design of supply networks; Inventory models for evaluation
          Ø Coordination (medium-term, tactical issues), e.g.,
               –    Centralized or decentralized control systems
               –    Inventory deployment
               –    Stock rationing
          Ø Coordination (short-term, operational issues)
               –    Inventory review, cost & service goals, control policies

                         Inventory Control

  State-of-the-art of the theory:
  • Single-stage systems with independent SKUs
     – well understood
  • Single-stage systems with dependent SKUs
     – fairly well understood
  • Multi-echelon systems
     – only partly understood and many problems
       are still in the research phase
  • Comprehensive supply chains
     – only in early research phase

                         Inventory Control

          Multi-echelon inventory systems
          Ø Fundamental interactions
               –   Lead-time effects of upstream shortages
               –   Upstream cost implications of downstream ordering policies
               –   Uncoordinated upstream requirement patterns

          Ø Other coordination/stock allocation issues
               –   Service measurement and safety stock interactions
               –   Rationing policies and partial shipments
               –   Emergency shipments and transshipments

          Ø Information and control systems

                                  Inventory Control

  General results for multi-echelon inventory systems with probabilistic demand
  Ø Serial systems
           –   No fixed ordering costs at downstream installations
                   • Imputed penalty costs                        (Clark & Scarf, 1960)
                   • Optimal policy: Base Stock / (R = 1, S i )   (Federgruen & Zipkin, 1984)

           –   Fixed ordering costs at downstream installations
                   • Optimal policy of (s i , ni Qi ) type        (Chen, 1997)
  Ø       Assembly systems
           –   Common in production environments
           –   Possible to decompose into a number of serial systems
                                                                  (Rosling, 1989)
  Ø       Distribution systems (Arborescent systems)
           –   No general optimality results
           –   Results for certain policy classes and system structures

                                    Inventory Control

               Status of system implementations:
               •     Numerous stand-alone dedicated software packages
                      – mostly for single-stage applications
               •     Simplistic methods in standard ERP systems
                      – often MRP oriented and deterministic for
                        multi-stage purposes
                      – often single-stage type to handle stochastic cases
               •     Developments for APS, etc.
                      – often network based and deterministically oriented for
                        tactical purposes
                      – particularly strong on seamless ICT/IS integration
                      – recent developments towards comprehensive
                        supply-chain issues
                        (See the examples to follow)

                        Inventory Control

          In general, there appears to be further needs for
              Ø Stochastic considerations (Davis, 1993, …)
              Ø Coordination issues
              Ø Combinations of the two
          …as well as for incorporating
            Ø Dynamic aspects           (Gavirneni & Tayur, 2001)
            Ø ‘New’ decision areas, e.g., auctions, spot markets,
              reverse flows, matching capacity…
                                        (Keskinocak & Tayur, 2001)
          in decision support systems for inventory control in SCM

                     Examples of
           Inventory Control Models in SCM

             Ø Global SCM at DEC
                (Arntzen et al, 1995)
             Ø Restructuring P&G’s supply chain”
                (Camm et al, 1997)
             Ø Decentralized multi-echelon inventory control
                (Andersson et al, 1998)
             Ø Teradyne’s service parts logistics system
                (Cohen et al, 1999)
             Ø Extended-enterprise SCM at IBM
                (Lin et al, 2000)
             Ø Risk Optimization in Oracle’s APS Suite
                (Oracle Corporation, 2000)

                          Global SCM at DEC
                               (Arntzen et al, 1995)

     Strategic decision support model and tool (GSCM)
           • Output:         Vendor-production-distribution network (with flows)
           • Scope:          Multiple products, facilities, echelons, technologies,
                             time periods, transportation modes
           • Constraints :   Demands, product structures, capacities, facilities,
                             local contents , offset trade, etc.
           • Objective:      Cost minimization (including lead-time effects)
           • Costs :         Fixed & variable production costs
                             Inventory holding costs
                             Distribution costs (multiple transportation modes,
                                                                taxes, duties, etc.)
           • Model type:     Mixed-integer linear programming (MILP) model
                             (Deterministic approach)

           • Reported savings: USD 100 million

                       Decentralized multi-echelon
                           inventory control
                                (Andersson et al, 1998)
          Ø Type of model
                –   Distribution system with one central warehouse and N retailers
                –   Continuous review, installation stock (Ri , Q) policies
                –   Decentralized decision making (autonomous installations)
                –   Limited information availability (‘actual lead times’)
          Ø Purpose
                –   Finding near optimal reorder points
                –   Minimizing expected holding and shortage costs
          Ø Results
                –   Incentive/penalty cost structure (‘fair’ sharing)
                –   Bounds for cost increase (approximation vs. optimal)

          Teradyne’s service-parts logistics system
                                        (Cohen et al, 1999)
          Ø   Company background
               –   Testing equipment for electronics assembly plants
               –   Complex service-parts repair and logistics system
                       •    Large number and variety of parts (costs and usage rates)
                       •    Geographic dispersion
                       •    Multiple classes of service
                       •    High customer shortage costs and service requirements
          Ø   Approaches
               –   Extensions of existing (basic) inventory models
               –   Team work
          Ø   Lessons learned
               –   Basic models effective both for
                   strategic analysis and operational control
               –   Applications require thorough understanding of both
                   theory and processes
               –   Simplicity enhances communication of insights and,
                   hence, facilitates implementation

          Teradyne’s service-parts logistics system
                                        (Cohen et al, 1999)

              Problem structure
              Ø Types of items
                   –       Consumables (high usage rates and low unit costs)
                   –       Repairables (return of defective items)

              Ø Types of repair service (costs and lead times)
                   –       Emergency (ES)
                   –       Regular (RR)

              Ø Logistics network
                   –       Central depot (CD) and repair center
                   –       Local centers (LC)

          Teradyne’s service-parts logistics system
                                   (Cohen et al, 1999)

          Suggested solutions:
          Ø Two types of models
               – Consumables
               – Repairables (More challenging task)

               – Consumables
                   • (R, S) control policy for CD (inexpensive, high-demand items)
                        – Normal distribution fit to empirical data
                        – Order fill-rate performance metric
                   • (S-1, S) control policy for LC’s (relatively low demand rate)
                        – Multi-echelon approximation
                        – Validation by simulation
                   • Result
                        – 90% reduction of late deliveries
                        – < 3% increase in inventory investment

          Teradyne’s service-parts logistics system
                                   (Cohen et al, 1999)

           Suggested solutions:
           •   Two types of models (cont’d)
                – Repairables
                    •   (S-1, S) inventory control policies
                    •   Single-location system as M/G/∞ queue
                    •   METRIC multi-echelon approach
                    •   Careful consideration of lead-time components

                    • Improvements compared to current policies
                    • Differentiated service policy dominates uniform policies
                    • Quantified benefits of lead-time reductions

               Extended-enterprise SCM at IBM
                                   (Lin et al, 2000)

          Ø Company background
              –   IBM Personal Systems Group (and other divisions)
              –   Global supply chain network
          Ø ‘Asset Management Tool’ (‘AMT’)
              –   Data Modeling, GUI, and Report Generator
              –   Optimization Engine
              –   Simulation Engine (based on SimProcess, CACI)
          Ø Results
              –   $750 M saved in 1998 (material costs and price-protection exp.)
              –   Evaluation of service levels related to forecast accuracy
                  from business partners
              –   Franz Edelman prize winner (INFORMS, 2000)
              –   Acquired by i2Technologies

               Extended-enterprise SCM at IBM
                                   (Lin et al, 2000)

          Ø Objective
              –   Trade-off between customer-service and inventory investment:
                  Minimize total expected inventory capital
                  s.t. customer-service targets
          Ø Modeling approach
              –   Multi-echelon network
              –   Base-stock inventory control policies (R = 1, S i )
              –   Each stocking location forms an M x/G/∞ queue
              –   Decomposition; Links through ‘actual lead time’ evaluations
              –   Rolling horizon heuristic
          Ø Solution procedures
              –   Gradient search algorithm
              –   Heuristic methods
              –   Interaction between Optimization and Simulation Engines


             Common characteristics of 3 examples
             • Decomposition approaches
             • Basic inventory control models for
               stochastic demand used as building blocks
             • Lead times between installations are
               essential for coordination

          Risk Optimization in Oracle’s APS Suite
                       (Oracle Corporation, 2000)

           Ø Tactical supply chain planning:
             Inventory planning
           Ø Make trade-offs between customer service levels
             and safety stock
           Ø Determine long-time procurement budgets
           Ø Make inventory postponement decisions
           Ø Identify strategic supply sources

                                (Source: Oracle Corporation)

          Risk Optimization in Oracle’s APS Suite
                         (Oracle Corporation, 2000)

          Characteristics and applicability
          Ø Model supply, demand, lead-time uncertainty
          Ø Stochastic optimization (prof. Infanger, Stanford Univ.)
          Ø Simulation capabilities

          Ø Calculate optimal time-phased safety stock deployment
          Ø Trade-off customer service and inventory levels
          Ø Recommend inventory postponement strategies

                                       (Source: Oracle Corporation)


To top