Integrated Planning – Module 2 1 Agenda • Forecasting, • Factors influencing Demand • Basic Demand Patterns • Basic Principles of Forecasting • Principles of Data Collection • Basic Forecasting Techniques, Seasonality • Sources & Types of Forecasting Errors Forecasting can be conducted at various levels Strategic Required for • Product life cycle • Long-term capacity planning • Capital asset/equipment/ human resource management Examples • Product line transitions • Annual volume out 3-5 years • Buy/build/lease decisions Financial Budgeting • Financial reporting • Working capital management • Production scheduling • Purchasing • Resource planning • Customer service management (product allocations) • Total annual/monthly volume • Projected product mix Operational • Weekly/monthly SKUlevel demand • Order size and frequency 3 Role of Forecasting in Supply Chain, • Basis for Strategic & Planning Decisions in SCM • Decisions needing Forecast as Base • Production – Scheduling -Inventory Control — Aggregate Planning – Purchasing • Marketing -Allocation of Sales-Force — Promotion Activities — New Product Launching Finance -Plant & Equipment Investment — Budgetary Planning • HR -Workforce Planning — Recruitment – Lay-Offs 3 Forecasting Impact Demand management Sales history/ orders Forecasting Demand planning Sales and operations planning Directly impacted by demand management Production management Aggregate planning Distribution management Inventory management Master production scheduling (MPS) BOM inventory routing Materials requirement planning (MRP) Distribution/transport network Shop floor scheduling and control Capacity requirements planning (CRP) Purchasing
Production 5 Higher forecast accuracy improves service levels at lower inventory Percent 100 99 98 97 96 95 0 94 Monthly average forecast error Excellent Far Poor 20% 40% 50% 3 1 2 Reducing forecast error will permit 1 Reduced inventory to a given service level 2 Increased service level for a given inventory level 3 Both reduced inventory and improved service level 2 4 6 Required average inventory Weeks 8 10 12 Forecasting error must be measured at different levels Forecasting error Percent Mean forecasting error Forecasting tips • Measure forecasting error as the ean absolute percent error (Forecast – actual sales) Forecast 60 50 40 30 20 10 0 -12 -10 -8 -6 -4 SKU/DC (12 oz. ketchup bottle in Dallas warehouse)* SKU level (12 oz. bottle) Brand level (ketchup) • Error of forecasting can be measured at various levels: product family, brand, SKU, SKU/DC and will improve at higher levels of consolidation • Frequency of measurement is usually monthly; however, best practitioners are doing weekly forecasts Product family level (all condiments) -2 Manufacturing lead time -0 • Measure bias as the mean percent error** (Forecast – actual sales) Forecast Required level of detail for planning 7 ** A consistently positive or negative bias indicates a tendency to over- or under-forecast which may be easily remedied Range of algorithms can be used COMMON FORECAST ALGORITHMS Model Simple Calculation/description • Last year plus percent • Last 3 months • Experiential smoothing • Seasonality with trend • Last year? s same period demand* increased by a flat percentage • Last 3-month moving average of demand • Last 12-month moving average with most recent 2-3 months more heavily weighted • Experiential smoothing with a seasonality factor that eights periods differently based on relative historical demand throughout the year • Regression • Incorporates variables other than historical demand (e. g. , price promotional activity) to best fit historical demand patterns • Time series • Real-time regression using POS data • Uses Fourier transforms to best fit historical demand patterns Complex • Modifies above models with changes in customer takeaway based on Nielsen; IRI data * Demand can be sales, shipments, orders depending on what works best and data available 8 Regression-based forecasting on high-promotion items
National cases shipped/week – ketchup example 90,000 80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 1 Fiscal week 10 20 30 40 5052 Actual peak shipment week is last week of fiscal month Post-promotion period Off-invoice promotions Key business drivers • Off-invoice promotions – Before a summer holiday – Without holiday • Promotion month end week • Postpromotion period Fiscal year 1 Fiscal year 2 Fiscal year 3 Higher peaks prior to Memorial Day and Labor Day 9 Forecasting implementation requires three success factors Tools and methodologies • “Single-point” forecast to manage (i. e. consistency across functioning) • Skills to balance art and science of forecasting Process • Driven by analytics, supported by market events • Explicit reconciliation steps Accountability • Accountability for both forecast error and inventory; need to balance trade-off • Rigorous measurement and tracking 10 Characteristics of demand Sources of demand – Customers – Spare parts – Promotions – Intra-company – Test samples – Others… ? All the sources of demand must be identified. Characteristics of demand Factors influencing demand – General business and economic conditions – Competitive factors – Market trends – Firm? own plans – Government regulations – Technology changes – Others… Characteristics of demand Components of demand – Trend – Seasonality – Random variation – Cyclical variation 40 35 30 25 20 15 10 5 0 2002 2003 Q1 Q2 2004 Q3 Q4 2005 6 Characteristics of Demand Trend Seasonal Demand Time 7 Characteristics of Demand Dynamic Stable Average Demand Time 8 Characteristics of demand Demand Patterns – Stable versus Dynamic > Stable demand has certain general pattern over time > Dynamic demand tends to be erratic – Independent versus Dependent > Demand for an item unrelated to demand for other items.
This is independent demand. > Demand that is directly related to derived from the bill of material structure of other items or end items. This is dependent demand. ? Only Independent demand needs to be forecasted. Dependent demand can be calculated. 9 Characteristics of demand Level of planning and forecast contents Forecast Business plan Market direction Time Frame 2 to 10 years Sales and operations Product lines and 1 to 2 years families End item and option Months Master production schedule 10 Characteristics of demand Why Forecast? Before making plans, an estimate must be made of what conditions will exist over some future period • Most firms cannot wait until orders are actually received before they start to plan what to produce • Manufacturers must anticipate future demand and plan to provide the capacity and resources to meet the demand • Firms that make standard products need to have salable goods immediately available / with shorter delivery time • Firms that MTO, must have labor and equipment to meet demand 11 Working without Forecast DemandForecasting Model 12 Principles of Forecasting and Data Collection
Forecasts.. – Are rarely 100% accurate over time – Should include an estimate of error – Are more accurate for product lines and families – Are more accurate for nearer periods of time While collecting data.. – Record data in terms needed for the forecast – Record circumstances relating to the data – Record demand separately for different customer groups 13 Forecasting Techniques Classification: – Quantitative Techniques – Qualitative Techniques – Intrinsic Techniques – Extrinsic Techniques – Short-range Techniques – Long-range Techniques 14 Qualitative Techniques Are based on intuition and informed opinion • Tend to be subjective • Are used for business planning and forecasting for new products • Are used for medium-term to long-term forecasting 15 Quantitative Techniques • Based on historical data usually available in the company • Assume future will repeat past 16 Extrinsic Techniques • Based on external indicators • Useful in forecasting total company demand or demand for families of products 17 Forecasting Techniques Moving Average: (Quantitative, Intrinsic) 3-period moving average Period 1 2 3 4 5 6 7 8 9 10 11 Period -3 -2 -1 Weightage 0. 0. 2 0. 7 Demand 265 240 295 265 310 285 304 312 328 299 267 267 290 287 300 300 315 313 281 269 300 288 301 308 322 306 Simple Weighted 18 Forecasting Techniques Moving Average: (Quantitative, Intrinsic) • Lags the actual sales. More the number of previous periods included, more is the lag • Can be used to filter out random variation • If a trend exists, it is hard to detect • Calculations become cumbersome when dealing with many time periods. More data storage required 19 Problem 1 • Over the past three months, the demand for a product has been 255,219 & 231.
Calculate the three month moving average forecast for month 4 • If the actual demand in month 4 is 228,calculate the forecast for month 5 Answer Moving Average Demand for 3 months= (255+219+231)/3 = 705/3 = 235 Moving Average for fourth month= (219+231+228)/3 =678/3 =226 Forecast for month 5 is 226 20 Forecasting Techniques Exponential Smoothing : (Quantitative, Intrinsic) Period Demand Forecast ( FT ) alpha (T) ( DT ) DT 1 2 3 4 5 6 7 8 9 10 11 190 160 220 200 300 240 270 200 290 275 305 0. 1 alpha= 0. 1 180 181 179 183 185 196 201 208 207 215 221 0. 5 alpha= 0. 180 185 173 196 198 249 245 257 229 259 267 0. 9 alpha= 0. 9 180 189 163 214 201 290 245 268 207 282 276 (FT+1) = FT + alpha (DT – FT) Forecasting Techniques Exponential Smoothing: (Quantitative, Intrinsic) • A type of moving average • Routine method for updating item forecasts • Satisfactory for short range forecasting • Can detect trends, but will lag them • Calculation and data requirements are manageable • Easy to „tune? 22 Problem 3 If the forecast for February was 122 and actual demand was 135,what would be forecast for March if smoothing constant is 0. 5, with exponential smoothing techniques. Answer In Exponential smoothing, forecast is calculated by formula (FT+1) = FT + alpha (DT – FT) = 122 + 0. 15( 135-122) = 122 + 1. 95 = 123. 95 say 124 23 Seasonality Key concepts: – Seasonality is variation in demand based on the season. – Seasonality may be annual, monthly, or even daily! – „Seasonal Index? is a measure of seasonal variation. Period average sales – Seasonal Index = Average sales for all periods – For forecasting purpose, de-seasonalized data is required. Seasonality Illustration:
Month Year1 Year2 Year3 Monthly Average 11. 00 12. 33 33. 33 48. 67 54. 67 56. 00 31. 33 19. 00 20. 33 17. 00 50. 33 49. 33 403. 33 Seasonal Index 0. 327 0. 367 0. 992 1. 448 1. 626 1. 666 0. 932 0. 565 0. 605 0. 506 1. 498 1. 468 12 33. 6 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total 10 13 33 45 53 57 33 20 19 18 46 48 395 12 13 38 54 56 56 27 18 22 18 50 53 417 11 11 29 47 55 55 34 19 20 15 55 47 398 Period average sales Seasonal Index = Average sales for all periods Average Sales for all months = 25 Seasonality Forecasting with Seasonality: Historical data is influenced by seasonality; hence can? t be used „as-it-is? for forecasting – Following steps are necessary: # Deseasonalize historical data # Forecast deseasonalized demand (Baseline Forecast) # Calculate the seasonal forecast by applying the Seasonal Index to the base forecast. 26 Problem 4 Month January February March April May June July Average Demand 30 50 85 110 125 245 255 Seasonal Index Forecast August September October November December 135 110 90 60 30 27 Month January February March April May June July August
Monthly Demand 30 50 85 110 125 245 255 135 Seasonal Index 0. 27 0. 45 0. 77 1. 00 1. 13 2. 22 2. 31 1. 22 New Av. Demand 166. 67 166. 67 166. 67 166. 67 166. 67 166. 67 166. 67 166. 67 Forecast 45. 28 75. 47 128. 30 166. 04 188. 68 369. 81 384. 90 203. 77 September October November December Total Average Sales for Month= 110 90 60 30 1325 110. 42 1. 00 0. 82 0. 54 0. 27 166. 67 166. 67 166. 67 166. 67 166. 04 135. 85 90. 57 45. 28 2000. 00 28 Tracking the Forecast Limitations of forecasts: – For several reasons, forecasts tend to go wrong. – We need methods to know ow good the forecasting method is. – „Tracking? is the process of comparing actual demand with the forecast – Forecast Error is the difference between actual demand and forecast demand – Error can occur in two ways: # Bias # Random Variation 29 Bias Bias exist when the cumulative Actual Demand varies from Cumulative Forecast Month Forecast Actual Monthly 1 2 3 100 100 100 Cumulative 100 200 300 Monthly 110 125 120 Cumulative 110 235 355 4 5 6 Total 100 100 100 600 400 500 600 125 130 110 720 480 610 720 30 Bias FORECAST ACTUAL DEMAND MONTHS 31 Random Variation
In a period actual demand will vary against average demand based on Demand pattern Month Forecast Actual Variation (Error) 5 -6 -2 4 3 -4 0 1 2 3 4 5 6 Total 100 100 100 100 100 100 600 105 94 98 104 103 96 600 32 Random Variation FORECAST 105 104 103 100 98 94 96 ACTUAL MONTHS 33 Tracking the Forecast Bias: – Bias is a systematic error in which the actual demand is consistently above or below the forecast demand – When bias is noticed, forecasting method should be changed to improve the forecast accuracy – For a unbiased forecasting method, the Cumulative Sum of Errors (CSE) will be zero 34
Tracking the Forecast Bias: (Illustration) Period 1 2 3 4 5 6 7 8 9 10 Total Forecast (F) 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 10000 Actual Sales 1200 1000 800 900 1400 1200 1100 700 1000 900 10200 Error 200 0 -200 -100 400 200 100 -300 0 -100 200 Interpretation: The bias (Average CSE) indicates that the there is an underforecast / positive bias of 20 per period. Cumulative Sum of Errors (CSE) 35 Forecast Error Measurement • Mean Absolute Deviation • Normal Distribution 36 Mean Absolute Deviation Forecast Error must be measured before it is used for planning or to revise the forecast • Mean Absolute Deviation ( MAD) commonly used for Error Measurement • Mean implies Average • Absolute means without reference to plus or minus • Deviation refers to the Error • MAD= Sum of Absolute Deviations Number of Observations in Earlier case, MAD = 5+6+2+4+3+4 = 24 = 4 6 6 37 Normal Distribution 1% -3 4% 15% 30% 30% 15% 4% -2 -1 0 1 2 3 1% +/- 1 MAD of the Average about 60% of the time +/- 2 MAD of the Average about 90% of the time +/- 3 MAD of the Average about 98% of the time 8 Use of MAD • Tracking Signal – to monitor Quality of Forecast • Tracking Signal= Algebraic Sum of Forecast Errors MAD • Past Six Month Consumption is – 105,110,103,105,107, and 115 ,where Forecast is 100 per month. • If MAD is 5 • Tracking Signal =(5+10+3+5+7+15)/5 = 45/5 =9 • Contingency Planning – Manufacturing Department can devise contingency plan for Capacity Utilization based on information regarding MAD of Forecast • Safety Stocks – Demand Variation is to be guarded by Safety Stocks 39 with Inventory Investment Decisions Tracking the Forecast
Mean Absolute Deviation (MAD): – MAD is a measure of random variation. – It measures the total error irrespective of the direction – For a normally distributed random variation, Standard Deviation (Sigma) = 1. 25*MAD – MAD can be used to determine: # Tracking Signal # Safety Stock 40 Tracking the Forecast Tracking Signal: – It is difficult to determine whether the variation is due to bias or random variation. – If the variation is due to random variation, the error will correct itself. – If the variation is due to bias, the forecasting method needs to be corrected. A tracking signal can be used to monitor the quality of the forecast. 42 Tracking the Forecast Tracking Signal: (Illustration) Period Forecast T 1 2 3 4 5 1000 1000 1000 1000 1000 1200 1000 800 900 1400 Sales W 1200 1000 1200 900 1400 Abs. Deviation T 200 0 200 100 400 W 200 0 200 100 400 T 200 200 0 -100 300 CSE W 200 200 400 300 700 CSE Tracking Signal = MAD 200 Tracking Signal (T) = = 1. 25 1200 Tracking Signal (W) = = 7. 5 160 160 6 7 8 9 1000 1000 1000 1000 1200 1100 700 1000 1200 1100 1300 1000 200 100 300 0 200 100 300 0 500 600 300 300 900 1000 1300 1300 10 1000 900 900 MAD= 100 160 100 160 200 1200 A tracking signal between +/- 4 means that the forecast is matching the actual data received. 43 Tracking the Forecast More about forecasts….. – Forecasts forecast average demand – Forecasts ignore random variations – Forecasting methods need to be continuously tracked and improved – Multiple forecasts should be avoided in a supply chain – If forecasting does not happen at right place, someone else is forced to do it – Certain operations are most affected by the forecast errors; postpone them as much as possible – The main aim of all the forecasting methods is to beat the naive forecast 44 Thank You 45
