Push-Pull: Strategic Thinking for Operational Excellence : Push-Pull: Strategic Thinking for Operational Excellence Yang Sun
Department of Industrial Engineering
Arizona State University Teaching Philosophy : IE@ASU 2 Teaching Philosophy Basics Intuition Synthesis Know How Know Why Major Source: Wally Hopp and Mark Spearman, Factory Physics, 2nd Ed., 2000 Basics : Basics Essential Context : IE@ASU 4 Essential Context The IDIB Portfolio
Sources: Lee Schwarz, Lecture Notes, 2003
Lee Schwarz, "A New Teaching Paradigm: The Information/Control/Buffer Portfolio", Production and Operations Management 7:2, pp. 125-131, 1998
Dan Shunk, “Knowledge Management”, Lecturer Notes. Variability Basics : IE@ASU 5 Variability Basics Variability is a fact of life. Increasing variability (always) degrades system performance.
The Bullwhip Effect (Volatility Amplification Law)
Think about Little’s Law!
Fruit Flies (Clockspeed Amplification Law)
Variability will be buffered by some combination of inventory, capacity, and time.
Hau Lee et al., Information distortion in a Supply Chain: The Bullwhip Effect, Management Science, 43(4), 1997; or The Bullwhip Effect in Supply Chains, Sloan Management Review 38(3), Spring1997
Charley Fine, CLOCKSPEED: Winning Industry Control in the Age of Temporary Advantage, 1998 Buffer Basics : IE@ASU 6 Buffer Basics Inventory
Definition: Lean = Minimal Buffer Cost
TOYOTA Lean Phases:
[Eliminate Direct Waste] ? (Value-Add)
[Substitute Capacity for Inventory Buffers] (Push -> Pull)
? [Reduce Variability] ? [Reduce Capacity Buffers] (Cont. Improv.) Sources: Wally Hopp, Supply Chain Sciences, 2005 Lessons from History : IE@ASU 7 Lessons from History A history of buzzwords
EOQ; MRP; MPR-II; ERP;
BPR; MES; APS;
Kanban; JIT; TQM;
What went wrong? Lessons from History (cont’d) : IE@ASU 8 Lessons from History (cont’d) Problems with traditional approaches:
Scientific Management has stressed math over realism
MRP is fundamentally flawed, in the basics, not the details
JIT is a collection of methods and slogans, not systems
Supply Chain/Manufacturing systems are large scale, complex, and varied.
No “technological silver bullet” can save us.
Continuous improvement is essential. Definition: Supply Network : IE@ASU 9 Definition: Supply Network A value-oriented network of processes and stockpoints that deliveries goods and services to customers. Sources: Wally Hopp, Supply Chain Sciences, 2005 Definition: Push/Pull Production System : IE@ASU 10 Definition: Push/Pull Production System Push Systems: schedule work releases based on demand. Pull Systems: authorize work releases based on system status. So, pull is not… : IE@ASU 11 So, pull is not… Kanban
Kanban is a special case of pull
ConWIP is a generalized pull concept
MRP with firm orders on MPS is make-to-order.
But it does not limit WIP and is therefore a push system.
Pull systems do replenish inventory voids.
But jobs can be associated with customer orders.
Toyota’s classic system made cars to forecasts.
Use of tact times or production smoothing often involves production without firm orders (and hence forecasts). The magic of pull… : IE@ASU 12 The magic of pull… Cycle Time (?t) ?
You don’t never make nothin’ and send it no place. Somebody has to come get it.
– Hall 1983
I dislike this definition.
The key is the WIP cap.
Why control the WIP?
Observability, Efficiency, and Robustness
Overcoming rigidity of pull WIP Exercise : IE@ASU 13 Exercise Are the following systems push or pull?
Kinko’s copy shop
Soda vending machine
“Pure” MRP system
Supermarket (goods on shelves)
Tandem line with finite interstation buffers
Runway at O’Hare during peak periods
Order entry server at Amazon.com Definition: Push/Pull Supply Chain : IE@ASU 14 Definition: Push/Pull Supply Chain A push supply chain makes production and distribution decisions based on forecasts (Build-to-stock)
A pull supply chain drives production and distribution by customer orders (Build/Assembly-to-Order)
Key concept: Location of the push/pull boundary (PPB) (strategic inventory point, inventory/order (I/O) interface)
Source: Simchi-Levi et al., Designing and Managing the Supply Chain, 2003 Push/pull Boundary Location : IE@ASU 15 Push/pull Boundary Location Real World Examples:
IBM PCB Case
GM Case (WSJ Oct. 21, 96, A1)
HP DeskJet Case (See Lee, Billington, and Carter, HP gains control of inventory and service through design for localization, Interfaces 23(4), 1993; Feitzinger and Lee, Mass Customization at HP: The Power of Postponement, HBR Jan-Feb, 1997)
Goal: eliminate entire portion of cycle time seen by customers by building to stock. (Need for responsiveness)
Basic Tradeoff: Responsiveness vs. Inventory (Time vs. Cost)
Levels: Product design (postponement) and process design (quick response mfg) Basic Takeaways : IE@ASU 16 Basic Takeaways Most systems are hybrid
Push/pull supply chain is a strategic design
Key: where the push/pull boundary is located
Lead time is the primary driving factor. (?t)
Push/pull production is a control policy
Push keywords: Ctrl Release ? Utilization
Pull keywords: Ctrl WIP ? Cycle Time (?t)
A pull thinking is always desired Intuition : Intuition The problem is choice : IE@ASU 18 The problem is choice Yes, but not only… Fisher’s Matrix : IE@ASU 19 Fisher’s Matrix Source: Marshall Fisher, “What is the right supply chain for your product”, Harvard Business Review, March-April 1997 Hau Lee’s Matrix : IE@ASU 20 Hau Lee’s Matrix Demand Uncertainty Low
(Functional Product) High
(Innovative Product) Low
(Stable Process) Low
(Functional Product) High
(Evolving Process) Supply Uncertainty Demand Uncertainty Low
(Functional Product) High
(Innovative Product) Low
(Stable Process) Low
(Functional Product) High
(Evolving Process) Supply Uncertainty Source: Hau Lee, “Aligning supply chain strategies with product uncertainties”, California Management Review, 44(3), 2002 Simchi-Levi’s Matrix : IE@ASU 21 Simchi-Levi’s Matrix Source: David Simchi-Levi et al., Designing and Managing the Supply Chain, 2003 Where to locate the PPB? : IE@ASU 22 Where to locate the PPB? Auto
Everything Material Assembly Parts
Goods Process Deliver CUSTOMERs Push Push Pull Pull Push Pull Push Pull Semiconductor Case Study : IE@ASU 23 Semiconductor Case Study Source: Yang Sun, Comparing Semiconductor Supply Chain Strategies under Demand Uncertainty and Process Variability, Master’s Thesis, ASU Forecasting and Demand Uncertainty : IE@ASU 24 Forecasting and Demand Uncertainty There is a confusion between two kinds of forecasting: ‘what can be sold (WCBS)’ and ‘what will be sold (WWBS)’ (Montgomery et al. Forecasting and Time Series Analysis, 1990). The former represents the possible market trends. The latter always represents the company’s capacity and budget constraint. Since capacity utilization is extremely important in semiconductor manufacturing, it is always the WWBS forecasts that triggers the production plan (push).
The semiconductor industry is always under stress: either ‘lack-for-sales’ (LFS) (WCBS < WWBS) or ‘lack-for-capacity’ (LFC) (WCBS > WWBS) (Shunk et al. Electronics Industry Drives of Intermediation and Disintermediation, submitted, 2005)
Note that huge demand uncertainty
EXISTS in the semiconductor
industry. WWBS WCBS WCBS (LFC) (LFS) Process Variability : IE@ASU 25 Process Variability Integrated into a cycle time distribution
Issues that can affect the variance of mfg cycle times: variable capacity, shortage of material, variable priorities in lot release, scheduling and dispatching, frequent machine breakdowns, operator error, etc.
Issues that can affect the variance of delivery time: globally distributed destination, regional traffic condition, variable 3PL/4PL, holding in custom Fab Probe Assembly Test Die
Goods Delivery Material CUSTOMER Front-end Back-end Delivery Performance Metrics: Cost and Service : IE@ASU 26 Performance Metrics: Cost and Service On-time delivery service is of critical importance in today’s semiconductor business, but companies are not doing very well today (case: Gateway penalized Intel by shifting business to AMD to blame Intel’s bad delivery service.)
Cost (per product sold) performance
= front-end mfg costs
+ back-end mfg costs
+ inventory (holding) costs
+ penalty costs based on tardiness
Delivery cost ignored
(Quality is a given in SC analysis.) Penalty L.T. Due The Simulation Model : IE@ASU 27 The Simulation Model Fab Probe Assembly Test Die
Goods Delivery Material CUSTOMER Front-end Back-end Delivery WWBS Order
WCBS Forecasts Push Pull
(Pull Strategy) Pull
Strategy) Performance DOE Factors : IE@ASU 28 DOE Factors fractional
factorial design Assume: two products, same family, assembled from common generic parent die A General Case Instance : IE@ASU 29 A General Case Instance And other assumptions Duarte, 2001 IC Knowledge, 2003 The ‘Global’ Experiment: Effects : IE@ASU 30 The ‘Global’ Experiment: Effects Since in simulation experiments almost all factors have none-zero effects, Sequential Bifurcation Analysis is suggested by Wan et al. 2003 (QSR Winner paper INFORMS Atlanta ‘03)
Group Screening: Factors are grouped as ‘Important’ and ‘Unimportant’
Step-Down: In each step, a group of factors are tested for importance Strategy Screened Factorial Effects : IE@ASU 31 Screened Factorial Effects Primary Factors: Due-date Lead Time and Penalty Weight (?t is the game)
Secondary Factors: Demand Pattern and Mfg Cycle Time Variability
Unimportant Factor: Final Product Logistics Time Variability
(Of course “costs” have significant effects, but do we need to analyze them?) Due-dates vs. Penalty Weights : IE@ASU 32 Due-dates vs. Penalty Weights Light Penalty Heavy Penalty Demand Pattern vs. Mfg C.T. Variability : IE@ASU 33 Demand Pattern vs. Mfg C.T. Variability $ Total Cost per wafer sold – Product A Low Variability High Variability Low Variability High Variability Low Variability High Variability Low Demand Mid Demand High Demand Low Demand Mid Demand High Demand $ Total Cost per wafer sold – Product A High Variability Low Variability High Variability Low Variability Low Variability High Variability High Variability medium due-date
and light penalty loose due-date
and heavy penalty The analytical results lead to a conceptual decision framework : IE@ASU 34 The analytical results lead to a conceptual decision framework Due-Date Lead Time Tight Medium Loose Importance of on-time delivery service Less Important Far More Important Push Pull Step-down to Layer Two Comparison This is Layer One Layer Two: Push-Pull Can be Appropriate : IE@ASU 35 Layer Two: Push-Pull Can be Appropriate * Manufacturing variability contains both front-end and back-end variability medium due-date + light penalty
loose due-date + heavy penalty Aggregate Demand Process Variability What else can be done? : IE@ASU 36 What else can be done? Pooling/Postponement
Hybrid Source: Alex Brown et al., Xilinx improves its semiconductor supply chain using product and process postponement, Interfaces, 30(4), 2000 Technology Involvement : IE@ASU 37 Technology Involvement Source: Joong-In Kim and Dan Shunk, working paper Intuition Takeaways : IE@ASU 38 Intuition Takeaways ?t is the name of the game. From the semiconductor case, lead time customers require and the perceived importance of on-time delivery are the driving factors.
We also need to understand not only the nature of the demands but that of the processes.
Supply Chain Visibility (both Demand Stream and Supply Stream) is important.
Implementation issues should be addressed.
Transition from push to pull needs tremendous cultural change and technological support. Synthesis -- Push-Pull It All Together : Synthesis -- Push-Pull It All Together A Customer-Driven Supply Chain Framework : IE@ASU 40 A Customer-Driven Supply Chain Framework Semiconductor SC example
Source: Yang Sun, Dan Shunk, John Fowler, Proceedings of INFORMS Annual Meeting, San Francisco, Nov. 2005 Info Flow Material Flow The Semiconductor Flow : IE@ASU 41 CUSTOMERS Die
Bank Wafer Fabrication (W/F) Assembly & Test (A/T) Configuration
& Shipment (C/S) The Semiconductor Flow Raw
Material Critical Decisions in the Semiconductor Supply Chain : IE@ASU 42 Critical Decisions in the Semiconductor Supply Chain Wafer Fabrication (W/F) Assembly & Test (A/T) Configuration
& Shipment (C/S) Source: Shunk et al., ASU DBR Survey, 2004 How are they made? : IE@ASU 43 How are they made? Wafer Fabrication (W/F) Assembly & Test (A/T) Configuration
& Shipment (C/S) Decision Technique Current Ideal? W/F A/T C/S 2004 Survey Result Logistic Estimation of Probabilities Inventory Management : IE@ASU 44 Inventory Management Key to Supply Chain Management
Deterministic model – adjust solution
- EOQ to compute order quantity, then add safety stock
EOQ Assumptions (not realistic)
Key Insight: There is a tradeoff between lot size and inventory
- news vendor model
- base stock and (Q,r) models
- (s,S) models
- Multi-echelon and network models Prioritizing and Releasing : IE@ASU 45 Prioritizing and Releasing There is sometimes confusion between the production planning domain and the shop floor control domain. We need to connect planning and execution.
The releasing function is key to Push-Pull. It connects supply chain planning and factory operation.
Supply Chain: Release by forecast vs. by order
Factory: MRP Push vs. Kanban/ConWIP Pull
Allocation is important for determining “who gets what”. Logistics : IE@ASU 46 Logistics Key to Supply Chain Management
Often performed by a 3PL or 4PL
Begin to contribute large portion to the GDP Within The Four Walls : IE@ASU 47 Within The Four Walls Capacity Release Scheduling Dispatching $$$$ $$$ $$ $ Recommended reading: John Fowler et al., Workload Control in the Semiconductor Industry, Production Planning & Control, 13(7), 2002 Shop floor ctrl
… Synthesis and Implementation : IE@ASU 48 Synthesis and Implementation The Strategic Importance of Details
The Practice Matter of Implementation
Creative alternative generation
Modeling and optimization
Communication and Teamwork Coordination and Collaboration : IE@ASU 49 Coordination and Collaboration Value of Info Sharing/SC Visibility
Coordinated Decision Making
Knowledge Sharing and Communities of Common Interests
VMI and CPFR
Remodeling the Supply Chains to pursue Supply Network Collaboration
Recommended Reading: Gérard Cochan, Matching Supply with Demand, 2005 Slide 50: IE@ASU 50 We think in generalities, we live in detail. –Alfred North Whitehead