Coping with Uncertainty: Reducing Bullwhip Behaviour in Global Supply Chains

Transcription

1 Coping with Uncertainty: Reducing Bullwhip Behaviour in Global Supply Chains Rachel Mason-Jones, and Denis R. Towill Logistics Systems Dynamics Group Cardiff University All supply chains suffer the effects of uncertainty. One of the most documented (and painful) symptoms is upstream order magnification known as the Bullwhip Effect. We believe that companies which cope best with uncertainty via an effective supply chain strategy are most likely to produce internationally competitive bottom-line performance. Although uncertainty takes many forms four key areas within a supply chain structure can represent these. These key segments are the supply, manufacturing, controls and demand. They may be combined to form the uncertainty circle. It is manifest that the uncertainty present within each of the quadrants must be reduced to maximise competitive advantage and hence market share. The paper therefore proposes a set of actions, which will reduce supply chain uncertainty caused by demand amplification (or the bullwhip effect ) irrespective of source. The methodology is shown to apply to a wide range of realworld supply chains. Acknowledgements Grateful thanks are extended to EPSRC who funded this research under Studentship Number in association with the PhD title Hierarchical Control of Supply Chain Dynamics. To protect market share and ensure survival companies have to meet f u t u re customer demand. A fore casting error can lead to inability to supply which not only results in a loss of sale at that moment in time but may impact future sales due to a consumer loss of confidence. Forecasting is a predictive which by its very nature carries an element of uncertainty. However f o recasting accuracy can be i m p roved by reducing the uncertainty experienced within the supply chain especially via lead time reduction. We believe that companies who understand and cope with uncertainty can optimise their forecasting potential and are in a better position to pro d u c e internationally competitive bottomline performance. Those companies who design business strategies which acknowledge the presence of uncertainty and provide mechanisms for pro-actively tackling it are rewarded by an opportunity to enable best practice ahead of competitors whose responses are purely re a c t i v e. Much uncertainty is system induced and magnified by the Bullwhip Effect as opposed to being introduced by the marketplace. Hence it is our experience that the best way to cope with uncertainty is to work hard to reduce it. The ultimate goal in our approach is the Seamless Supply Chain (SSC) wherein all players think and act as one (1). Thus the SSC obtains a greater market share to the benefit of all the players within the chain. To understand the problems posed by uncertainty, let us conr the operation of a typical company engaged in the Product Delivery Process (PDP) in which goods are supplied in response to an order from our immediate customer. But who our immediate customer actually is depends on our position within the chain. For example if we are an OEM, our immediate customer is usually a Dealer, but if we are a PCB manufacturer our customer is usually an electronic products s u b - a s s e m b l e r. As we shall see l a t e r, in the traditional supply Supply Chain Fo r u m An International Journal N w w w. s u p p l y c h a i n - f o r u m. c o m

2 Supply chain, uncertainty will be a function of how far upstream we are from the ultimate marketplace. The further upstream we are, the worse will be the Bullwhip Effect. There are various contributing factors to supply chain uncertainty, however the source of the Bullwhip effect quoted by Lee et al (2) is that of the amplification dynamics of the order distortion as it propagates upstream. Regardless of our position within the chain, the PDP uncertainty problem may be simplified and put into the generic format of Figure 1. Here a single echelon PDP is schown with our value added (which may be composed of many individual tasks) is directed by the controls. The company responds to our immediate customer (the ). In turn our stocks are replenished with materials, components, and sub-assemblies by various vendors (the Supply ). Our conred view is that reducing uncertainty is achieved by understanding and tackling the root causes inherent in each of the four areas in Figure 1. The purpose of this paper is to demonstrate that an holistic approach to Supply Chain Management (SCM) based on this generic model has a great deal to offer in achieving these goals using the Uncertainty Circle as a framework for improving performance. Shrinking the uncertainty circle Control systems Manufacturing (our value added) Product Delivery Process Material flow Information flow Fig. 1 - Single Generic Model of the Causes of Uncertainty in the Product Delivery Process. shows a typical situation in which all four are significant and approximately equal in importance. What does then management typically do to reduce uncertainty? Two decisions are normally taken, sometimes sequentially, sometimes concurrently. Firstly, we strive to improve the performance of the value added by reducing lead times and greatly improving quality levels via the application of Lean Thinking (4). Secondly, we work more closely with our suppliers. Preferably via a Partnership Sourcing Programme. We thereby expect to conrably improve supplier quality, reduce supplier lead times and experience much more consistent delivery patterns so that JIT production may be enabled as appropriate. Unfortunately companies which have taken these actions (often a conrable cost in money, time, and hassle) are often perplexed to find that substantial operational uncertainty still persists. The reason is quite obvious as can be seen from Fig. 2 (b), since we have only reduced two of the sources of uncertainty, leaving the controls Supply Planning and Control systems Manufacturing and demand problems untouched. Furthermore management is up against the law of diminishing returns here. There is thus little point in continuing to hammer our suppliers and our shop floor for even better performance particularly when there are big gains remaining to be made elsewhere in the system. But of course we have so far concentrated on the easy actions. It is manifestly much simpler to re-engineer the shop floor and chase/change our suppliers than attempt to persuade all players in the chain to seriously practice SCM! It is thus obvious that to deliver the further improvements shown via the Seamless Supply Chain in Fig. 2 (c) that we must be able to amply demonstrate the benefits accruing from successful SCM. This we shall do using both simulation and realworld supply chain results. The Bullwhip Effect in Traditional Supply Chains Forrester (5) first simulated the demand amplification characteristic exhibited by real-world supply chains over thirty years ago. More recently this phenomenon has become known as Bullwhip Effect due to the characteristic increasingly magnified and hence worsening behaviour observed upstream from the source disturbance. Burbidge (6) offered an explanation for this behaviour and called it Law of Industrial Dynamics. Hence the Bullwhip effect can be seen as a resultant of the following law: Significantly reduced uncertainty As we approach the seamless supply chain Fig. 2 - Shrinking the Supply Chain Uncertainty Circle the Key to Enchanced Performance. Control systems The major factors in supply chain uncertainty of Manufacturing (our value added ), Supply Side, Controls, and Side may be represented by the four fundamental segments in the PDP Uncertainty circle (3). Figure 2 (a) (A) Traditional supply chain (B) Improved supply and manufacturing improved via lean thinking Significantly reduced costs As we approach the seamless supply (C) Improved demand and control systems via information sharing and better systems design chain Supply Chain Forum An International Journal N

3 - If demand for products is transmitted along a series of inventories using stock control ordering, then the demand variation will increase with each transfer. So the demand amplification inherent in supply chains is system induced and is directly affected by both information and material delays in the chain and the feedback loops in the decision making. A real world example of the effect this has on the dynamic behaviour of a supply chain is schown in Figure 3. The retail supply chain example used was initially presented by Stalk and Hout (7) and it clearly shows the increasing magnification of order fluctuations as they proceed further upstream. At each company within a supply chain, order information is subject to delay, bias, and noise before transferring it onto their subsequent supplier. Far from cancelling each other out, these various sources of distortion of true marketplace demand is at best additive and at worst multiplicative in their effect. Hence any genuine marketplace uncertainty is greatly magnified as orders are transmitted upstream in the supply chain. Frequently this is compounded by players at all levels double guessing the true order pattern. Unfortunately their actions are usually out-of-phase with what is actually required to damp down excess magnification of the order waveform. The result is typified by the demand amplification pattern exhibited in Figure 3 where the upstream yarn manufacturer is having to cope with order fluctuations in excess of 40% greater than the downstream retailer. The Bullwhip Effect has been widely observed and researched to discover opportunities for good supply chain design. The practical realworld consensus is that if the demand can find a way to multiply then it will, (8). So, the further away from the end consumer a player is in the supply chain, the less he is aware of the true consumer demand and the more misled he is likely to be especially since there is usually geographical as well as time separation. In a traditional supply to (+/-) 40% (hence Bullwhip effect here = 2x2x2=8 times greater than marketplace variation) Yarn Maker chain the raw materials supply will respond to an increase in consumer demand maybe weeks, months or even years after the buying practice first showed itself in the marketplace. By this time the product may well be obsolete with vast quantities of stock written off at various points throughout the chain. Simulation models capable of answering what - if questions are a valuable tool in engineering the supply chain (9). But despite the fact Forrester made his observations in the early 1960 s the problems identified then still hold true in the majority of today s chains (10). The problem of demand amplification and its subsequent dynamic effects have not gone away despite the enormous amount of research and the flurry of supply chain improvement programs. Uncertainty still persists, but learning to constrain the effects by developing strategies that deal effectively with it are crucial supply chains activities to ensure they can actively compete in the marketplace. Before a supply chain can achieve this goal it needs to accept that uncertainty is a problem and target the root causes in each of the four quadrants of the Uncertainty Circle. Tackling Uncertainty in both Pipelines Reduces the Bullwhip Effect Jay Forrester first coined the phrase pipeline management to describe the continuing controlled flow of goods on demand using the analogy with the behaviour of a Direction of Amplification and Increasing Uncertainty Fabric Maker as the Wave form moves Upstream to (+/-) 20% Garment Maker to (+/-) 10% Information Pipeline (Upstream flow) Material Pipeline (Downstream flow) Hight Street Retailer to (+/-) 5% Fig 3 - Sequential Structure of a"traditional"supply Chain Generating the Bullwhip Effect in the Clothing Sector (Based on Description by Stalk and Hout, 1990) hydraulic system. This embraced the view that there is a continuous flow of materials through the supply chain rather than via a series of unconnected operations. Effective Supply Chain Management requires the development of seamless strategies which smooth the flow by removing barriers between companies. All supply chains have two distinct lead-time pipelines, the order information transfer pipeline moving upstream from point of sale to raw material supplier and secondly the product transfer downstream from raw material to customer. The information and material pipelines are clearly in figure 3. The uncertainty Circle (previously show in Figure 2) illustrates the four ket areas within a supply chain where uncertainty is present. Each of the four segments suffer from uncertainty in both the information and material pipelines. As previously stated many supply chains have targeted their material pipelines and achevied reduced uncertainty via, for example, optimised manufacturing es. However the information pipeline still has much potential in providing a great opportunity for uncertainty reduction. Even if a supply chain achevies maximum benefit from streamlined material flow its overall dynamic behaviour in comparison to the actual marketplace will still be subjected to demand amplification if the information pipeline is not similarly improved. Companies must understand that optimisation of material flow alone is not enough to achieve Customers Supply Chain Forum An International Journal N

4 Strategic review Agile manufacturing program Vision of Agile response Business inputs Feasibility study a sustained competitive advantage and the key differentiator may well be how effectively the order data is transferred upstream. One industrial example of the performance benefits that can be achevied by ensuring both information and material flows are optimised is shown below in figure 4. Figure 4 was originally presented by Towill (11) based on data from Goldman (12). As an example to show the type of opportunities available in the information pipeline Figure 4, originally presented by Fine (13) compares the actual marketplace sales and the order pattern imposed on the upstream machine tool manufacturer. At first sight the dynamics shown could be from two separate non-related companies as opposed to the reality of being part of the same supply chain. The dynamics of the information pipeline shown in Figure 5 clearly illustrate the Move to orientation Open up information flows Synchronise manufacture JIT information availability People regarded as core asset Proposed actions Implement -action On-time shipments doubled Quality up by 75% Cycle times down by 25-75% Work In Progress down by 50% Output metrics Fig. 4 - Input-output diagram summarising the impact of streamlined flow controls on the performance of a company. (Towill (11) Fig 5. - The Supply Chain Bullwhip Effect in the Machine Tool industry(12) 80% 60% 40% Bullwhip effect and hence the uncertainty that builds up within the supply chain resulting in wild swings in capacity requirements. The way to attenuate the Bullwhip effect is to reduce uncertainty caused by distorted data and hence effectively damp down the demand amplification behaviour. Replacement of a product is initiated by a consumer buy therefore PDP is activated by the marketplace sales information (3). Thus speed of information transference is crucial to an effective supply chain. Unlike production delays, which are reliant on improving hardware technological es, the order information pipeline can in theory be literally at the speed of light. Not surprisingly, EDI has become essential to ensure a competitive market position for many industries. The importance of information has long been recognised in the change, and is described by Hammer and Champy (14) as a key-enabler for BPR. Speed of information transfer has therefore been recognised as a key commodity that can, if handled properly, become a competitive advantage. However it is the quality of the information not the quantity of data which is the key enabler. Many companies have discovered this the hard way having implemented IT they have received none of the expected payback (15). The danger is that blindly bolting IT on to existing information systems without ensure an optimised will only transfer bad data faster - it will not improve the quality of the information. To ensure improved a dynamic behaviour and hence reduced uncertainty, information needs to be treated as a commodity whose pipeline strategy can be re-designed. To ensure competitive advantage and maximised bottom line performance supply chains should seek strategies that limit the effect of uncertainty in both the material and information pipeline across all four segments of the Uncertainty Circle. Targeting Material and Information flow shrinks the circle There are a variety of tools which can be utilised to improve supply chain performance across all four quadrants of the Uncertainty circle. Some key ones are summarised in Figure 6. As can be seen the sources of uncertainty are in both the material and the information pipelines. Fig 6 - Examples of improvement strategies to reduce Uncertainty in each Quadrant. Planning and Control Systems % Change year to year 20% 0% -20% Postponement of product customization Partnering Schemes Information Good Decision Support System Single communication channel between players Push/Pull approach -40% -60% Rationalize Vendor Base Partnering Scheme Information Sharing Eliminate waste (Muda) Consistency of product Consistency of times Year % change GDP % change vehicle production index % change net neworders machine toot industry Supply Manufacturing Process Supply Chain Forum An International Journal N

5 Uncertainty reduction strategies within the material flow pipeline are well developed and have been applied in many industries, one of the most notable techniques is via Lean Thinking (4). The best way to reduce uncertainty within the order information pipeline is to ensure everyone in the supply chain gets the most up-to-date and useful information. As market sales data is the catalyst information for the whole supply chain it is the source of the undistorted data. Therefore by directly feeding each level of the supply chain with the market sales data actively tackles the order uncertainty from the point of source. This enrichment is achievable with a point of sales link at each level in the supply chain. So rather than each link making an order decision based purely on the internal chain sales data it can now make an informed judgement based on what the end consumer is actually buying. A standard way of establishing benchmarks for judging the effectiveness of competitive supply chain re-designs is via the use of simulation models (16). We use a generic model for this purpose, which is a tried and tested representation of good industrial practice. Each echelon is represented by a Decision Support System (DSS) which takes account of orders received, Work In Progress (WIP), stock policy, and any forecasts available. The echelons are then coupled together in the way which best describes current supply chain strategy. One example of such a generic model depicting the behaviour of an electronic products supply chain is described Supply Chain Design Strategy Datum Design Information pipeline redesign only Material pipeline redesign only Both pipelines re-designed Bullwhip Performance Measures by Berry et al. (17), which was used to benchmark various real-world strategies against the baseline case. We find that for synthesis purposes it is realistic to use a step demand to simulate a shock event at the marketplace. The Bullwhip waveform propagation upstream is then observed. The reason for this is that the step response provides an extremely rich picture of the supply chain dynamics. Hence the approach adopted is to make a preliminary selection of the best strategy on the basis of the step response, and then verify the choice by simulating random and seasonal disturbances (18). One benchmarked output from such a simulation study is shown in Table 1 in which the scenarii have been selected to demonstrate the impact of lead-time reduction and information enrichment on the Bullwhip Effect. Four strategies were simulated to illustrate the effect of adopting improvement programmes. The first was a datum design, which simulated a supply chain before any improvement. The second simulated design adopting a strategy, which targeted the information pipeline alone by enabling all players sight of the actual marketplace sales. The third design simulates an improved material pipeline by speeding the flow so the chain can be more responsive. The fourth strategy simulates a total Uncertainty Circle approach where both the material and information pipelines are optimised. Table 1 - Benchmarking Supply Chain Bullwhip Performance. Performance Improvement Benchmarking: Shrinking the Uncertainty Circle Peak Value Peak time recovery Stock depletion Stock recovery Overall Bullwhip Performance Benchmark = Best Bullwhip performance, = Worst Bullwhip performance In order to benchmark the various strategies a methodology for ranking the competitive designs was needed. Hence for each benchmark the best and worst design in terms of Bullwhip behaviour were highlighted on a linear scale and were ranked accordingly with four stars (best) and one star (worst). The remaining designs were then ranked on this scale by reference to the best and worst designs as observed during the simulation. If the difference between the two designs is regarded as insignificant, both designs are given the same star rating. From Table 1 it can be seen that reducing the uncertainty in each of the pipelines has a positive effect on the overall performance of the supply chain. The effect the material and information pipelines have on supply chain behaviour is seen as significantly different. Information is crucial to the control of the supply chain whereas material flow directly effects the speed of response. However the dramatic benefits available from adopting a strategy which reduces uncertainty in both the material and information pipelines is clearly evident in Table1. Not only has the Bullwhip effect been reduced in the supply chain but equally importantly the response times have also greatly improved the ordering function and the stock levels. For this particular supply chain configuration it is clearly advantageous to reduce all material flow lead times and operate an information enrichment policy. Industrial Examples of Shrinking the Uncertainty Circle One good example of market sales information being used within the consumer market is the business link between Wal-Mart Stores (USA) and Procter & Gamble, one of its main suppliers. Basically Wal-Mart dismantled the barriers to information sharing and opened its consumer information to its suppliers for their tactical and strategic usage. This was a revolutionary step resulting from the new Wal-Mart ideology that it did not matter what competitors might thereby learn Supply Chain Forum An International Journal N

6 SUPPLY CHAIN PERFORMANCE METRIC Reduction in production demand amplification variability Reduction in central warehouse stock variability PRODUCT CODE A B C D E F -45% -26% -18% -25% -14% -46% -34% -31% -35% +15% -27% -36% (a) Improvement in production orders and central warehouse stock variability The Uncertainty Circle is a useful concept in greatly improving the performance of real-world supply chains to the benefit of all players therein. Whereas the Supply Side and Manufacturing Process segments are essentially under the direct control of the business and may be tackled using principles such as Lean Thinking, the Process Planning and Control and Side require action to be taken externally if significant improvements are to be achieved. Both the Process Controls and uncertainties in product delivery may be substantially shrunk via the ready transparent availability of undistorted marketplace data. To obtain the necessary collaboration businesses may find the Partnership Agreements are good enabling mechanisms. Our particular approach to shrinking the Uncertainty Circle has been demonstrated via simulation modelling and real-world BPR experiences. Only if strategies, which tackle the uncertainty present in both the information and material pipelines in all four segments of the Uncertainty Circle, are adopted can maximum opportunity be realised. REFERENCES YEAR % TOTAL CHANGE about the business so long as the Wal-Mart relationship with their suppliers grew stronger thus resulting in a better service for customers (19). Procter & Gamble now take the consumer information from Wal-Mart at point of sale and decide how much stock to deliver to the stores so as to ensure consumer demands are satisfied. It is Procter & Gamble responsibility to keep Wal-Mart shelves full through a Vendor Managed Inventory (VMI) strategy. This strategy has a double benefit in terms of shrinking the size of the Uncertainty Circle. Firstly it reduces uncertainty by eliminating both material and information flow delays; secondly it reduces uncertainty yet further by eliminating a decision point within the chain. Another real-world supply chain application of the principles of shrinking the Uncertainty Circle has been reported by Towill and McCullen. (20). It concerns the reengineering achieved in a mechanical precision product global supply chain via their rapid response and IT integration programmes. Actions taken and implemented include manufacturing lead time reduction (manufacturing quadrant); linking factories direct to customer demands (demand quadrant); more frequent and more rapid planning (planning and control quadrant); streamlined acquisition and distribution (supply quadrant); and a new Decision Support System (again the control quadrant) to exploit reduced time delays. Table II summaries the performance improvements achieved by the company through reducing uncertainty in each of the four quadrants of the uncertainty circle % -29% -35% -45% Table II - Improved Supply Chain Dynamic Performance Observed Following Implementation of Agile Manufacturing Programme (Towill and McCullen (11)) The recorded results presented in Table II include the downward trend in (reducing by 45% over a five year period), and up to 46% reduction in the Bullwhip effect across the sampled products. Unfortunately the data does not enable the credit to be apportioned between rapid response elements and IT Integration elements of these concurrent re-engineering programmes but the combined benefits are indeed impressive. They also confirm the effectiveness of our approach to shrinking the Uncertainty Circle as a means of reducing the Bullwhip Effect. CONCLUSION 1. Towill, D R The Seamless Supply Chain - The Predators Advantage International Journal of the Technology of Management Vol. 13, No 1, 1997, pp Lee H.L, Padmanabhan V and Whang S. (1997) Information Distortion in a Supply Chain: The Bullwhip Effect Management Science Vol. 43, No. 4, pp Mason-Jones R and Towill, D R (1998) Shrinking the Supply Chain Uncertainty Circle Control Vol. 24,No. 7,pp Womack J., Jones D.T. (1996) Lean Thinking, Simon and Schuster 5. Forrester J. W, (1960), Industrial Dynamics. MIT Press Cambridge MA 6. Burbidge, J L. (1989) Production Flow Analysis Oxford University Press, Oxford. UK. 7. Stalk. G, Hout T Competing Against Time Free Press Towill D R Supply Chain Dynamics - The Change Engineering Challenge of the Mid 1990s Prod I.Mech.E in Eng Man Vol. 206, 1992, pp Andraski J.C. (1994), Foundations for a Succesful Continuous Replenishment Programme Int. J. Logistics Management., Vol 5, No 1, pp Balle M, (1995) The Business Process Re- Engineering Action Kit Kogan Page. London. 11. Towill D.R, (1999) Simplicity Wins, Twelve Rules for Designing Effective Supply Chains Control March Issue, pp Goldman S.L, Nagel R.N Preiss (1995) Agile Comptitors and Virtual Organisation Van Nosbrand Reinbold, New York. 13. Fine C., (1998) Clockspeed: Winning Industry Control in the Age of Temporary Advantage. Sloan School of Management, M.I.T. 14. Hammer M and Champy J (1993) Re-engineering the Corporation Harper-Collins. New York. 15. Davenport T. (1994), Saving IT s soul: human centred information management Harvard Business Review. March-April pp Ackere A.V, Larsen E. R, Morecroft J. D.W. (1993), Systems Thinking and Business Process Redesign: An Application to the Beer Game European Management Journal, Vol 11, No 4, pp Berry D, Naim M.M and Towill D.R, (1995), Business Process Re-Engineering an Electronic Products Supply Chain IEE Proc-Sci Meas. Technol. Vol 142, No 5, September. 18. Mason-Jones R. K, Naim M.M, Towill D. R The Impact of Pipeline Control on Supply Chain Dynamics International Journal of Logistics Management, Vol. 8, No 2, 1997,pp Johansson H. J, McHugh P, Pendlebury A.J, Wheeler III W.A. Business Process Reegineering. (Wiley), Towill D.R and McCullen P. (1999) The Impact of an Agile Manufacturing Program on Supply Chain Dynamics Accepted for publication in International Journal of Logistics Management, December Issue. Supply Chain Forum An International Journal N