18 nov - forecasting and bullwhip effect
TRANSCRIPT
![Page 1: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/1.jpg)
ì BA 244: Supply Chain Management
![Page 2: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/2.jpg)
Internal Supply Chain
Procurement Produc,on Storage Distribu,on Sales and Marke,ng
Forecas,ng
![Page 3: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/3.jpg)
Demand Management
ì Balances customer requirements with the capabili,es of the supply chain (Lambert, 2008)
ì Process within an organiza,on to “tailor its capacity, to meet varia)ons in demand, or to manage the level of demand using marke)ng or SCM strategies” (CIPS)
![Page 4: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/4.jpg)
Successful Application
ì Use detailed POS data to match the rate of produc,on to demand: forecast
ì Need for established process for receiving, storing, and using POS data from retailers (Lawrie, 2007 b:2)
![Page 5: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/5.jpg)
Successful Application
1. Define relevant data to manage demand, followed by systema,c and accurate recording of this data
2. Synchronize demand with supply
3. Long term planning
4. Strategically assess promo,onal ac,vity and its impact on demand
Taylor and Fearne (2006)
![Page 6: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/6.jpg)
Hints and Tips
ì Collabora,ve demand forecas,ng ì firms reach a consensus, both internally and with
their chain partners on the expected level, ,ming, mix and loca,on of demand (Lawrie, 2007b)
ì Use pricing and promo,ons to s,mulate demand (Lawrie et al., 2007b)
ì Monitor sales against forecasts
![Page 7: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/7.jpg)
Forecasting
ì Encompasses techniques employed to systema)cally analyze data and informa)on in an aXempt to predict future paXerns, trends or performance (Lysons and Farrington, 2006)
ì Underlying basis or all business decisions ì Produc,on ì Inventory ì Personnel ì Facili,es ì Budget
![Page 8: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/8.jpg)
Steps in Forecasting
1. Select the items to be forecast
2. Determine the ,me horizon of the forecast
3. Select the forecas,ng model
4. Gather data
5. Make the forecast
6. Validate and implement results
![Page 9: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/9.jpg)
Forecasting Approaches
Qualita,ve Approach
When liXle data exist
Involves intui,on, experience
e.g. new technologies, new products
Quan,ta,ve Approach
When stable and historical data exist
Involves mathema,cal techniques
e.g. current technology, exis,ng products
![Page 10: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/10.jpg)
Forecasting Approaches
ì Which approach did they use to forecast the iPad?
![Page 11: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/11.jpg)
Overview of Quantitative Approaches
Quan,ta,ve Forecas,ng
Time Series Models
Moving Average
Exponen,al Smoothing (EWAM)
Trend Projec,on
Associa,ve Models
Linear Regression
![Page 12: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/12.jpg)
What is a Time Series?
ì Evenly spaced numerical data ì Regular ,me periods
ì Forecast based only on past values ì Assumes that factors influencing past and present
will con,nue to influence in the future
![Page 13: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/13.jpg)
Overview of Quantitative Approaches
Quan,ta,ve Forecas,ng
Time Series Models
Moving Average
Exponen,al Smoothing (EWAM)
Trend Projec,on
Associa,ve Models
Linear Regression
![Page 14: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/14.jpg)
Forecasting Demand
ì Moving Average ì How to best smooth out fluctua,ons
![Page 15: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/15.jpg)
Moving Average
PAST FUTURE
![Page 16: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/16.jpg)
Moving Average
PAST FUTURE
![Page 17: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/17.jpg)
Moving Average
PAST FUTURE
![Page 18: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/18.jpg)
Observations?
ì What is the trend? ì Upward ì Downward
ì Which data is more smooth? ì Demand ì 3-‐week ì 6-‐week
![Page 19: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/19.jpg)
However
ì Which data is more relevant? ì Older? ì Most Recent?
ì Example: Philippine Popula,on for 2017 ì 2010: 150 M? ì 2016: 200 M?
![Page 20: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/20.jpg)
Forecasting Demand
ì Exponen,ally Weighted Average Method (EWAM) ì Since the older the demand data, the less relevant ì Adds weights, with more weight to more recent
data ì Weights must add to 1 or 100%
![Page 21: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/21.jpg)
Exponentially Weighted Average Method (EWAM)
Week Demand, pcs Weight Weighted Qty Weighted FC 1 650 0.2 130 2 678 0.3 203 3 720 0.5 360 4 ? 693
![Page 22: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/22.jpg)
Overview of Quantitative Approaches
Quan,ta,ve Forecas,ng
Time Series Models
Moving Average
Exponen,al Smoothing (EWAM)
Trend Projec,on
Associa,ve Models
Linear Regression
![Page 23: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/23.jpg)
Trend Projection
![Page 24: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/24.jpg)
Trend Projection
ì Y = a + bx ì Y is the forecast ì a is the intercept ì b is the slope
ì Extrapola,on
![Page 25: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/25.jpg)
Overview of Quantitative Approaches
Quan,ta,ve Forecas,ng
Time Series Models
Moving Average
Exponen,al Smoothing (EWAM)
Trend Projec,on
Associa,ve Models
Linear Regression
![Page 26: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/26.jpg)
Regression
ì Demand is expressed as a func,on of an independent variable, not ,me
ì Demand is forecasted by plugging values of the independent variable
ì e.g. sokdrink demand as a func,on of temperature
ì e.g. product demand as a func,on of promo budget
ì Key is to have a logical rela,onship between variables
![Page 27: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/27.jpg)
Regression Demand for Burger Steak
Meal, pc Demand for Chickenjoy
Meal, pc Marke,ng Budget for Burger Steak, PHP ,00
1 37 11,190 2 53 15,930 10 83 25,020 9 64 19,200 11 48 14,400 13 58 17,490 17 64 19,140 16 66 19,740 20 53 15,900 18 80 23,970 19 77 23,100 24 71 21,150 21 75 22,410 28 60 18,030 26 79 23,730 30 97 29,130 36 73 22,020 41 81 24,420 40 70 21,120 42 62 18,450 48 90 27,060
![Page 28: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/28.jpg)
Regression
Variable Demand for Burger Steak Meal
Independent Variables
Marke,ng Budget for Burger Steak 29.948 **
Demand for Chickenjoy Meal -‐18.685 *
![Page 29: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/29.jpg)
Forecasting Time Horizon
ì it is difficult to be as accurate the further into the future they go; there are poten,al risks associated with longer horizons
ì technological products with short life cycles can only be forecasted a few months into the future ì vs. furniture product forecas,ng that can be done
for years ahead, since furniture products have a longer life cycle (Boyer and Verma, 2010)
![Page 30: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/30.jpg)
Forecasting Time Horizon
![Page 31: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/31.jpg)
Performance Monitoring
ì forecast accuracy should be monitored and its assump)ons, techniques and validity of data revisited when the actual outcomes differ considerably from those predicted (Lysons and Farrington, 2006)
ì Goodness-‐of-‐fit tests
![Page 32: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/32.jpg)
ì Bullwhip Effect
![Page 33: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/33.jpg)
History
ì First described by Forrester in 1958 and has been experienced since the 1960s
ì Term was first used in the management circles of Proctor & Gamble, when in the 1980s the company experienced extensive demand amplifica,ons for Pampers
![Page 34: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/34.jpg)
Definition
ì Demand distor)on that travels upstream in the supply chain due to the variance of orders which may be larger than that of sales (Lee and Billington, 1992)
![Page 35: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/35.jpg)
Cause
ì Inventory is oken a subs)tute for informa)on, as any kind of uncertainty is covered by inventory. However, adding in safety stocks can send out false signals and encourage suppliers to also compensate for uncertainty by similarly building in safety stocks
![Page 36: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/36.jpg)
Effect
ì accumula)on of inventory at the manufacturer's end, which further increases supply chain costs to the company (Sucky, 2009)
ì stockholding and obsolescence costs
![Page 37: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/37.jpg)
Mitigations
ì Reduced lead ,mes
ì Shared knowledge with suppliers and customers to beXer gauge demand; Provide each stage of the supply chain with complete access to customer demand informa)on
ì use of technology to speed communica)ons and improve response ,me
![Page 38: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/38.jpg)
Steps to Successful Application
ì Improve communica)on and informa,on flow along the supply chain
ì Improve data forecas)ng (e.g. determining product demand from actual data entered into POS computer systems will improve sales forecast accuracy)
ì Work with firms upstream and downstream in the supply chain
ì Order products up and down the supply chain in smaller increments, thus reducing the )me between orders and allowing for )mely informa)on to be available
Fransoo and Wouters (2000)
![Page 39: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/39.jpg)
Other Tips
ì Eliminate variability of demand caused by unplanned promo,onal ac,vi,es at the retailers' end (Towill et al.,1996)
![Page 40: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/40.jpg)
Case Studies
ì The Barilla S.p.A. case was one of the first published studies to empirically support and provide illustra,ons of the issues resul,ng from the bullwhip phenomena
ì One of the major pasta producers in Italy, offered special price discounts to customers
ì ordered full truckload quan,,es
ì Resulted in spiky and erra,c customer order paXerns
ì As a result, supply chain costs outstripped the benefits from full truckload transporta,on
(Barilla S.p.A., HBS Case 9-‐694-‐04)
![Page 41: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/41.jpg)
Case Studies
ì HewleX Packard printers
ì When examining the actual sales at a major reseller, execu,ves found that there were some normal fluctua,ons over ,me
ì However, when they examined the orders from the reseller, they observed much bigger swings
ì Moreover, the orders from the printer division to the company's integrated circuit division had even greater fluctua,ons
(Kuper and Branvold, 2000)
![Page 42: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/42.jpg)
Internal Supply Chain
Procurement Produc,on Storage Distribu,on Sales and Marke,ng
Forecas,ng
Corporate Spend Management
![Page 43: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/43.jpg)
Internal Supply Chain
ì How much to produce or purchase ì Demand Planning
ì Which one to produce or purchase first? ì Corporate Spend Management
![Page 44: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/44.jpg)
Pareto Analysis
ì 80/20 Rule
ì 80% of spend being directed towards just 20% of the suppliers
ì Cri,cal Few vs. Trivial Many
![Page 45: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/45.jpg)
ì Pareto Analysis
![Page 46: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/46.jpg)
HOMEWORK
ì READ!
ì Han, K., et. al. 2012. Value Cocrea,on and Wealth Spillover in Open Innova,on Alliances. MIS Quarterly. 36: 291-‐315.
![Page 47: 18 Nov - Forecasting and Bullwhip Effect](https://reader034.vdocument.in/reader034/viewer/2022042604/5876c6261a28ab6d5a8b5e0b/html5/thumbnails/47.jpg)
ì End