International Journal of ISSN 0974-2107 Systems and Technologies Vol.4, No.1, pp 83-93 IJST KLEF 2010 Case Study on Forecasting, Bull-Whip Effect in A Supply Chain T.V.S. Raghavendra 1, Prof. A. Rama Krishna Rao 2 & P.V.Chalapathi 3 1,3 Associate Professors, Department of Industrial & Production Engineering, K.L. University, Vijayawada. E-mail: rvdgmc@gmail.com 2 Department Of. Mechanical Engineering, S.V. University, Tirupati Abstract: An important phenomenon often observed in supply chain is bullwhip effect, means demand variability increases as we move up in the supply chain. In this paper we quantify this effect for simple, two stage supply chain. We demonstrate the application of forecasting technique by retailer can cause a bullwhip effect and contrast these results with the increase in variability. We considered winter s model of forecasting and tracking signal for a demand- order relation. We then discuss several important managerial insights that can be drawn from the research. Key words: Supply chain, forecasting, bull-whip effect, tracking signal. 1.0 INTRODUCTION There are some thumb rules for demand forecasting: first, forecasts are always inaccurate. Creating forecasts at the lower levels and then grouping them accordingly for planning purposes is vital to a healthy process. Forecasts are only the starting point for the planning process forecasts help provide a basis for further refinements Actual consumer demand may be very different from the order stream. Each member of the supply chain observes the demand patterns of its customers and in turn produces a set of demands on its suppliers. But the decisions made in forecasting, setting inventory targets, lot sizing and purchasing act to transform (or distort) the demand picture. The further a company is upstream in the supply chain (that is, the further it is from the consumer), the more distorted is the order stream relative to consumer demand. This phenomenon is also known as the Forrester effect or bull-whip effect. It is important to see the meaning of the bull-whip effect both in downstream and upstream of the value chain This leads to a demand curve with steeper and steeper peaks and downs and with less and less reliability the further up the party is in the value chain. In the upper stream of the value chain the parties are forced to take extreme actions to survive the peaks only to find out that the demand was exaggerated. The total cost of the value chain is increased heavily and the reliability and timelines of the deliveries has suffered. Also the fragmented organizations in companies have led to atomistic 83
T.V.S.Raghavendra Considerations, i.e. sub-optimization of business activities, which cause the bullwhip effect to occur internally in the company. The multiplied effect of the intra-organizational and cross-enterprise suboptimization and non-collaborative, non-synchronized, individual processes leads to the bullwhip curve. In other words that means that the internal forecasting process is operating well, and the company has a common plan or forecast at both ends. The final aim is to have centralized demand information one forecast. The four material flow principles, which can be used to reduce the bullwhip effect, are control system, time compression, information transparency and echelon elimination. In this paper, considering a typical supply chain consisting of plant, retailer and customer. effect, Future demand using one of the forecasting techniques, tracking signal at retailers are calculated and related. Further an attempt is made to find a relation Fig 1 : Bull whip effect in Supply chain between bullwhip effect and tracking signal for this case. The graphical representations above show the bull whip effect between two supply chain partners. It can be seen that distributor orders to the factory experience demand fluctuate far more drastically than the retail demand. Over time as the distributor builds inventory and fulfills orders, it communicates very different demand levels to the upstream factory by the order amount it requests. This becomes more complicated the farther up the supply chain we go. 84
Case Study on Forecasting.. 1. CAUSES OF THE BULL WHIP EFFECT The identified four major causes of the bull whip effect are as follows 1. Demand forecast updating. 2. Order batching. 3. Price fluctuation. 4. ning and shortage gaming. 2. DEMAND FORECAST UPDATING Every company in a supply chain usually does product forecasting for its production scheduling, capacity planning, inventory control and materials requirement planning. Forecasting is often based on the order history from the company s immediate customers. When a down stream operation places an order, the upstream manager processes the piece of information as a signal about the future product demand. Based on this signal the upstream manager readjusts his or her demand forecasts and in turn the orders placed with suppliers of the upstream operation. 3. CONSEQUENCES OF THE BULL WHIP EFFECT: Lower revenues. Stock-outs and backlogs mean lost sales, as customers take their business elsewhere. Higher costs High carrying cost. Distributors need to expedite orders (at higher shipping expenses). Manufactures need to adjust jobs (at higher setups and changeover expenses, higher labor expenses for overtime, perhaps even higher materials expenses for scarce components.) All entities in the supply chain must also invest heavily in outsized facilities (plants, warehouses) to handle peaks in demand, resulting in alternating under or over-utilization. Worse quality. Quirky, unplanned changes in production and delivery schedules disrupt and subvert control processes, begetting diverse quality problems that prove costly to rectify. Poorer service. Irregular, unpredictable production and delivery schedules also lengthen lead time, causing delay and customer dissatisfaction. 85
T.V.S.Raghavendra 4. SUPPLY CHAIN STRUCTURE Fig 1 : Demand-order variability It consists of three members, a manufacturer, a retailer and a consumer. In each period t, the retailer observes his inventory position and places an order quantity to the manufacturer. After the order is placed, the retailer observes and fills customer s demand for that period. Any unfilled demand is backlogged. There is a fixed lead time between the time, an order is placed by the retailer and when it is received by the retailer. Manufacturer Wholesaler Retailer Customer 5. QUANTIFICATION OF BULLWHIP EFFECT One measurement of bullwhip effect is the ratio of output order rate manufacturer) to the input order rate (consumer demand) ( retailers order to Effect = In general the ratio can be calculated as effect = The ordering quantity can be expressed as below Let assume that the retailer follows a simple order up-to inventory policy in which order-up-to point is estimated from the observed demand as S t = D L t + Z * σ L t 86
Case Study on Forecasting.. L D t σ L t : demand over L periods : standard deviation of the demand over L periods Z : constant chosen 6. DESIGN OF EXPERIMENT The purpose of the design of experiment is to analyze the impact of a) Smoothening parameters (α, β, γ ), b) Trend component c) Strength of seasonality On the quantification of bullwhip effect and on relation between bullwhip effect & tracking signal. 6.1 Dependent variable The dependent variable is ratio. It is the ratio of quantity ordered by the retailer to the manufacturer to order (or) demand observed by the retailer. 6.2 Smoothening parameters Alpha ( 0.7 to 0.9) Beta ( to 0.05) Gamma ( to 0.05) 6.3 Strength of Seasonality Low seasonality Medium seasonality High seasonality 7. FORECASTING TECHNIQUE It is assumed that the retailer utilizes the Winter s (triple exponential smoothening) model to forecast the demand. Table 1 shows the forecasted values for 3 years F t+1 = (L t + T t ) * I t+1-m Where L t = α ( D t / I t-m ) + ( 1-α ) ( L t-1 + T t-1 ) T t = β ( L t L t-1 ) + ( 1-β ) T t-1 I t = γ (D t / L t ) + (1-γ ) I t-m m is the month number F t+1 is the Forecast component at period t+1 L t is the Level component of demand period t 87
T.V.S.Raghavendra T t is the trend component of demand period t I t+1-m is the Seasonality index for the same period in previous year 8. FUTURE SCOPE OF RESEARCH Assess the impact of bullwhip effect of performance measures of supply chain (i.e. total chain cost, total cost of members, service level of chain and service level of chain members). Investigate other techniques of time series for seasonality & Lead time effect on Bull-whip effect. 8. CONCLUSION 1. From the analysis, the is minimum for Lower value of Alpha, lower values of Beta and small higher values of Gamma as shown in Table 2.and Fig 3, 4 & 5. 2. The months of Year is divided into Three Seasons namely Higher, Medium and Lower. An analysis of the relation between Effect with a range of values of Alpha, Beta and Gamma with variations of Seasonality is done and inferences are as followed as shown in Fig. 6, 7 & 8. 2.1.1 During High Seasonality, to minimize the Effect, the values of Alpha should be at the lowest as shown in Table 3 (a). 2.1.2. During High Seasonality, to minimize the Effect, the values of Beta should be at the lowest as shown in Table.3 (b) 2.1.3. During High Seasonality, to minimize the Effect, the values of Gamma should be at the lowest as shown in Table.3 (c) 2.2.1. During Medium Seasonality, to minimize the Effect, the values of Alpha should be at the lower value near to optimal value as shown in Table 3 (a). 2.2.2. During Medium Seasonality, to minimize the Effect, the values of Beta should be at the higher value as shown in Table 3 (b). 2.2.3. During Medium Seasonality, to minimize the Effect, the values of Gamma should be at the lowest value as shown in Table 3 (c). 2.3.1. During Low Seasonality, to minimize the Effect, the values of Alpha should be at the optimal value as shown in Table 3 (a) 88
Case Study on Forecasting.. 2.3.2. During Low Seasonality, to minimize the Effect, the values of Beta should be at the optimal value as shown in Table 3 (b). 2.3.3. During Low Seasonality, to minimize the Effect, the values of Gamma should be at the higher value as shown in Table 3 (c) 11.0 REFERENCES 1. Chopra, S, Meindyl P, 2001. Supply chain Management, Pentice-Hall New Jersey. 2. Chen F, Z.Drezner, J.K. Ryan and D. Simchi Levi, 2000a. Quantifying the bullwhip effect in a simple supply chain: The impact of forecasting, lead times and information, Management Science, 46, 436-443. 3. Lee H, PadmanabhanV and Whang S, 1997a. Information Distortion in a Supply Chain: The Effect, Management science, 43, 546-558. Table. 1 : Retailers statistics for 3 years Month 2007-08 2008-09 2009-10 Actual Actual Actual Demand Demand Demand Demand Demand Demand April 4608140 ---- 5090301 4629078 5035224 4704420 May 4668783 4713925 5211502 5112166 5007126 5084537 June 4655509 4675824 5138866 5184401 5077607 5019118 July 4709680 4863044 5436783 5374174 5133432 5281736 August 4851126 4853842 5442617 5546746 5750024 5284081 September 4801511 4550226 4765322 5129794 5436401 5288090 October 4755305 4756385 4767970 4866084 5540531 5417812 November 4776948 4889109 4940490 4923678 5318657 5666122 December 4860493 4815168 4799055 4946957 5406163 5413988 January 4864709 4733130 4481643 4716293 5154063 5277106 February 4455077 4613680 4403733 4327448 5226119 4946186 March 4883501 5064160 5028594 4939668 5085430 5814570 ratio 1.248698892 0.982596542 1.752948417 89
T.V.S.Raghavendra Variable Alpha Beta Gamma Level 0.7 0.76 0.8 0.9 0.01 0.05 0.01 0.05 forecasted value Subset 1.09685E+11 1.12601E+11 1.14488E+11 1.19167E+11 1.12054E+11 1.12601E+11 1.17174E+11 1.12606E+11 1.12601E+11 1.12563E+11 1.055767597 1.083834495 1.101999812 1.147035436 1.078567107 1.083834495 1.12785511 1.083882955 1.083834495 1.083468986 Original variance value 2 : Retailers Statistics ( Varied α, β, γ values Vs ) 1.12601E+11 1.083834495 Ta ble Analyzing the effect of Range Values of α, β, γ on Effect The values of Alpha varies in the range ( 0.7 to 0.9) while values of Beta and Gamma are kept at a constant values as 0.01 each. The values of Beta varies in the range ( to 0.05) while values of Alpha and Gamma are kept at a constant values as 0.76 and 0.01 respectively. The values of Gamma varies in the range ( to 0.05) while values of Alpha and Beta are kept at a constant values as 0.76 and 0.01 respectively. Fig. 3 : Alpha values with 90
Case Study on Forecasting.. Fig. 4 : Beta values with Fig. 5 : Gamma values with Table. 3 (a). Alpha values, Seasonality & for Retailer s Statistics The values of Alpha vary in the range ( 0.7 to 0.9), while values of Beta and Gamma are kept at a constant values as 0.01 each. Variable Level Seasonality Alpha 0.7 value sub set 168412621.3 1.626841479 0.76 162903565.4 1.573624798 High 0.8 157445468.1 1.520900369 0.9 141134560.9 1.363339374 0.7 207415667.3 1.614918451 0.76 194385825.3 1.513469353 Medium 0.8 190673172.9 1.484562998 0.9 192014984.2 1.495010213 0.7 2686821490.3 13.33172969 0.76 2366701048.2 11.74332524 0.8 Low 2453583809.1 12.17442851 0.9 2686821490.1 13.33172969 Original variance value Actual Demand 162903565.4 1.573624798 103521224.1 194385825.3 1.513469353 128437239.2 2366701048 11.74332524 201535851.2 91
T.V.S.Raghavendra (b) Beta values, Seasonality & for Retailer s Statistics The values of Beta vary in the range ( to 0.5), while values of Alpha and Gamma are kept at a constant values as 0.76 and 0.01 respectfully. Variable Level Seasonality Beta value sub set 157755186.1 1.523892202 0.01 High 162903565.4 1.573624798 0.05 161747027.3 1.562452805 0.01 Medium 201851132.9 194385825.3 1.571593521 1.513469353 0.05 145617637.1 1.133764927 2581337722.4 12.80833017 0.01 Low 2385862269.2 11.83840123 0.05 2559072474,1 12.69785231 Original variance value Actual Demand 162903565.4 1.573624798 103521224.1 194385825.3 1.513469353 128437239.2 2385862269 11.83840123 201535851.2 (c) Gamma values, Seasonality & for Retailers Statistics The values of Gamma varies in the range ( to 0.5), while values of Alpha and Beta are kept at a constant values as 0.76 and 0.01 respectfully. Variable Level Seasonality Gamma value sub set 157755186.1 1.523892202 0.01 High 162903565.4 1.573624798 0.05 161747027.5 1.562452805 0.01 Medium 194277406.2 194385825.3 1.51262521 1.513469353 0.05 195274182.5 1.520386017 2396117688 11.88928756 0.01 Low 2385862269 11.83840123 0.05 2305300894 11.43866404 Original variance value Actual Demand 162903565.4 1.573624798 103521224.1 194385825.3 1.513469353 128437239.2 2385862269 11.83840123 201535851.2 92
Case Study on Forecasting.. Fig. 6 : High Season Fig. 7 : Medium Season Fig. 8 : Low Season 93