Demand Information Distortion and Bullwhip Effect

Transcription

1 Demand Information Distortion and Bullwhip Effect Chopra: Chap. 17 Assignment 2 is released.

2

3 Lessons of the Game Such oscillations are common Bullwhip effect (demand distortion) Everyone blames others - but problem is with the structure

4 Bullwhip Effect Manufacturers Regional Local Local Local Distributors Wholesalers Retailers Customers Bullwhip effect: increased demand variability up the SC

5

6 Cambell Chicken Soup

7 Bullwhip effect in the US PC supply chain Changes in demand 80% 60% 40% Semiconductor Equipment 20% PC 0% -20% Semiconductor -40% Annual percentage changes in demand (in $s) at three levels of the semiconductor supply chain: personal computers, semiconductors and semiconductor manufacturing equipment.

8 HP Laser: L Series 4L 5L 4L 5L HP- Shipment Wholesaler-seethru

9 Sell out Std Dev 5L Comp USA OfficeMax Office Micro Electronic Depot Elek Tek Best Buy PC Warehouse Order Std Dev Staple 12000

10 Causes for Poor SC Performance Demand uncertainty ( how to cope with it?) Product variety ( -- ) Safety stock Better forecast. Better plan. Information distortion along the SC -- bullwhip ( -- )

11 Curses of Bullwhip Effect Curses

12 Frequency = 10 Svc Level =.95 P(Stockout) =.05 = = 350 x =? X Safety Stock = x -

13 Causes of Bullwhip Effect Key causes Demand forecasts update (by different parties) information distortion Leadtimes Price promotion - forward buying Order synchronization Batch ordering practice Shortage Gaming Psychological effect? Not in the game

14 Demand forecast updating by Intuition Each location forecasts demand to determine shifts in the demand process How should a firm respond to a high demand obs? Is this a signal of higher future demand or just random variation in current demand? If the firm s inventory is low, hedge by assuming this signals higher future demand, i.e., order more than usual How should a firm respond to a low demand obs? If the firm s inventory is high, be more conservative and wait to see if demand has really shifted, i.e., no order now Rational reactions at one level propagate up the SC

15 Forecast Updating SC: An Example Order-up-to Level Period t t-1 t-2 t-3 t-4 t-5 Demand Forecast Order Upto Order q Assumptions: retailer uses D t-1 to forecast future demand D t, as F t = D t-1 ; order-up-to-level 2 F t. No Safety stock?

16 Impact of Forecasting on BE The BE is due, in part, to the need to forecast demand & hold safety stock Moving ave and exponential smoothing are bad The fancier the method, the worse the BE Smoother demand forecasts can reduce the bullwhip effect (MA & ES methods) The longer the leadtime, the higher the BE Centralised information sig reduces the BE

17 q T, L=0 Retailer D T If all updating their forecasts, variability amplifies exponentially Forecast Updating SC Mkt Demand : and Thus, ordering For and follows for period for q t a period quantity : F T S Var (q D Var (qt ) Var (D ) example, Cov( D t t D Var (q naive S t -1: ) z t 2 : (, ). z t T S T T-1 t 1. t : 2 2 forecasting D Var (D [2 t 1 zf t 1, D S - Cov (D ) 0, ) / Var (D ) 1.5 T S 2 Var (q 2 t-1 t t ) t 2 t Retailer :Order up - tolevel z F z F z( D t-1 t 1 - D z 2 t 2 method : t 1 t-2 D ) t-1 z D t 2, D z D t 1 ) t-2 z 1.5 : t-2 Assume: extra inventory can be returned without any cost in [-, + ] 2 )] z 2

18 Avoiding Demand Forecast Updates BE resulted from the chain effect along the SC Repetitive multiple forecast updating Share demand information so that every one can obs demand shifts without distortions: Demand forecasts should be based on final sales to consumers

19 Bullwhip can occur within a firm Volvo Green Cars Sales We need to promote and get rid of these green cars Production All green cars are sold out, time for replenishement

20 Avoiding Demand Forecast Updates Channel Alignment VMI - vendor managed inventory scheme Consumer direct Discount for information sharing, including plan of promotion activities Operational Efficiency Leadtime reduction Echelon-based inventory control

21 Order Synchronization Synchronized ordering occurs when retailers tend to order at the same time: end of the week orders beginning of the month orders end of the quarter orders

22

23 U n its Order batching Retailers may be required to order in integer multiples of some batch size, e.g., case quantities, pallet quantities, full truck load, etc The graph shows simulated daily consumer demand (solid 30 line) and supplier demand (squares) when retailers order 20 in batches of 15 units, i.e., every 15 th demand a retailer 10 orders one batch from the supplier that contains 15 units T ime (e a c h p e rio d e q u a ls o n e d a y )

24 Order batching solutions Smaller min order quantity (lower Q), so retailers order more frequently Unsynchronize retailer order intervals Retailers may order every T periods Min batch size Q=1, so no min order Q restriction Retailers are placed on balanced schedules s.t. average demand per period is held constant e.g., 100 identical retailers and T=5 implies 20 retailers may order each period

25 Trade Promotion Why trade promotion? Consequences of trade promotion?

26 D e c Ja n Fe b M a r A pr M a y Jun Jul A ug Se p O c t N ov C a se s C a se s Trade promotions and forward buying Supplier gives retailer a temporary discount, called a trade promotion. Retailer purchases enough to satisfy demand until the next trade promotion. Example: Campbell s Chicken Noodle Soup over a one year period: Total shipments and consumption One retailer s buy 7000 Shipm e nts C onsum ption T im e (w e e ks)

27 On-hand inventory (units) Time (measured in weeks)

28 Shortage game Retailers submit orders for delivery in a future period Supplier might not be able to fill all orders He might not get enough components His production yield might not be as high as expected Phantom orders Reatilers order more than they think they need to make sure they get a good allocation if demand is high or if capacity is tight

29 When is shortage game likely? Supplier allows retailers to cancel order or accepts returns High retailer profit margin, i.e., costly to not have goods Retailer demand expectations positively correlated(i.e., if one retailer has high demand expectation, the other retailers probably do too.) Retailer competition (if retailer A takes more inventory, retailer B has less to sell) Capacity is expensive, so the supplier will not build unlimited cap

30 Classic Bullwhip Effect: Semiconductor Industry, 1995 Perception: Demand for semiconductors would have a tremendous increase Result: Customers, worried about a supply shortage, tripled their orders Reality: Semiconductor companies scrambled to meet demand, realized information was inflated and suffered huge losses

31 Shortage gaming: bad and good Bad: supplier can t use initial orders to forecast demand, so it builds the wrong level of capacity allocation among retailers is poor: some retailers get more than they need, others are starved Good: Reduces idle cap., assuming the retailers actually take and sell the product

32 How to stop phantom ordering Don t let retailers cancel orders Don t offer retailers generous return policies Share cap. And inventory data prevent false scares Prioritize retailers (customers, e.g., by past sales)