Description
forecasting is required, principles of forecasting, factors that influence a forecast, different forecasting methods, time series forecasting.
Forecasting
Forecasting is the art and science of predicting future events Institute of business forecasting
Why Forecast?
Why Forecast?
Forecasts empower people because their use implies that we can modify variables now to alter (or be prepared for) the future. A prediction is an invitation to introduce change into a system. There are several assumptions about forecasting: There is no way to state what the future will be with complete certainty. Regardless of the methods that we use there will always be an element of uncertainty until the forecast horizon has come to pass. There will always be blind spots in forecasts. We cannot, for example, forecast completely new technologies for which there are no existing paradigms. Providing forecasts to policy-makers will help them formulate social policy. The new social policy, in turn, will affect the future, thus changing the accuracy of the forecast.
1.
2.
3.
Why Forecast?
Lead times require that decisions be made in advance of uncertain events. ? Forecasting is an important requirement for all strategic and planning decisions in a supply chain. ? Forecasts of product demand, materials, labor, financing are an important inputs to scheduling, acquiring resources, and determining resource requirements.
?
Demand forecast and Supply Chain Management
Customer’s order is the starting point of initiation of a firm’s business activities, including logistics and supply chain management. If a firm is in a make-to-stock production strategy, SCM is required to plan the level of its activities, such as procurement, transportation, warehousing etc. drugs If a firm in in make-to-order production strategy, SCM is required to plan capacities – machine, labour etc – an aircraft
Demand forecast and Supply Chain Management
In both cases a Supply Chain manager needs a forecast.
Mature markets have a more stable and predictable demand forecast – milk, most food products, items of personal consumption
For unpredictable demand, for example mobile phone models, demand forecast is very difficult – error can cause under or over production and supply.
What is a Forecast Error
?
Demand forecast estimate for a product A is between 30 and 70 units for a period, with a mean of 50 and mean deviation of 20. For a product B the demand forecast estimate is between 40 and 60 units, with the same mean 50 and mean deviation of 10. The 2 firms will be required to organise their logistics differently, if the market demand as per estimates is to be met; with different cost implications. The difference between forecast and actual demand is the forecast error.
?
Principles of Forecasting
?
? ?
?
Forecasts are almost always wrong. Every forecast should include an estimate of the forecast error. The greater the degree of aggregation, the more accurate the forecast. Forecast for total mobile phone demand for India will be more accurate then forecast for sale of Nokia’s N72 model in Pune. Long-term forecasts are usually less accurate than short-term forecasts. Forecasting sale of cold drinks for today and tomorrow is more predictable than for the next week – weather influence, other local factors
Factors that influence a forecast
Demand is influenced by a variety of factors. We must identify each factor that influences demand of the product – » state of economy, » competitors’ activities, » discounts and promotions, » seasonal factors, » substitutes and complimentary products, » and so on.
Factors that influence a forecast
Subjective factors must also be used in correcting demand forecast. A quantitative technique based forecast for umbrellas would not be able to factor the impact of current season’s monsoon activity. Same for woolens, cold drinks and other similar products.
Basic Forecasting Approach
?
?
?
Understand the forecasting objective. What decisions will be made from the forecasts? What parties in the supply chain will be affected by the decision. Integrate demand planning and forecasting. All planning activities within the supply chain that will use the forecast or influence demand should be linked. Collaborative forecast improves response capability to match supply and demand. Identify factors that influence the demand forecast. Is demand growing or declining? Are there relationship (complementary or substitution) between products?
Forecasting Approach (cont.)
?
?
?
Understand and Identify customer segments. Customer demand can be separately forecast for different segments based on service requirements, volume, order frequency, volatility, etc. Demand behaviour for electricity by industries and commercial establishments in festive season differs Determine the appropriate forecasting technique. Typically, using a combination of the different techniques is of the the most effective approach. New products may use judgemental method Establish performance and error measures. Forecasts need to be monitored for their accuracy and timeliness. Eliminate causal effects of a one time promotion.
Forecasting Horizons.
Short Term (0 to 3 months): for inventory management and scheduling. ? Medium Term (3 months to 2 years): for production planning, purchasing, and distribution. ? Long Term (2 years and more): for capacity planning, facility location, and strategic planning.
?
Forecasting Methods
?
?
Qualitative methods are subjective in nature since they rely on human judgment and opinion. Quantitative methods use mathematical or simulation models based on historical demand or relationships between variables.
Selecting a forecast method:
Some Qualitative Methods
?
Jury of Executive Opinion (opinions of a small group of high-level managers is pooled).
Sales Force Composite (aggregation of salespersons estimate of sales in their territory).
?
?Delphi Method (a forecasting group uses a staff to prepare, distribute, collect, and summarize a series of questionnaires and survey results from geographically dispersed respondents, whose judgements are valued).
Some Qualitative Methods
Simulation Methods imitate the consumer choices that give rise to demand to arrive at a forecast. ? Market Research Method (solicit input from customers or potential customers regarding future purchasing plans).
?
Quantitative Forecast Methods
?
?
Time Series Methods use historical data extrapolated into the future. They are best suited for stable environments. Moving averages, exponential smoothing methods, time series decomposition, and Box-Jenkins Methods. Causal Methods assume demand is highly correlated with certain environmental factors (indicators). Correlation methods, regression models, and econometric models.
Time Series Demand Model
?
Observed Demand = Systematic Component + Random Component.
Systematic Component measures the expected value of demand and consists of: ?Level: the current deseasonalized demand. ?Trend: the rate of growth or decline in demand. ?Seasonality: the regular periodic oscillation in demand. Random Component is that part of demand that follows no discernable or predictable pattern.The random component is estimated by the forecast error (forecast – actual demand).
?
?
Time Series Forecasting
?
Static
?Assume estimates of level, trend, and seasonality do not vary as new data is observed.
?
Adaptive
?Update forecast as new data becomes available
Time Series Forecasting
Static ? Adaptive
?
» Moving average » Single exponential smoothing » Trend-adjusted exponential smoothing (Holt’s) » Trend & Seasonal adjusted exponential smoothing (Winter’s)
Static Forecasting (steps)
1. 2. 3.
4.
5.
Determine periodicity of seasonality (even or odd impacts calculation of deseasonalized demand?) Deseasonalize data = Average of consecutive perods demand for all seasons in a cycle Find the equation of the trend line a. Simple linear regression – D = L + Tt - intercept and slope b. Independent variable (period) c. Dependent variable (deseasonalized data) Estimate seasonalized factors a. Per period = observed demand / deseasonalized dem b. Index (Averages) = Average of all period factors Forecast
Find the equation of the line
?
Use simple regression
» Excel: (Tools/Data Analysis/Regression) » Dependent variable
is deseasonalized demand » Independent variable (x) is period t » y= intercept + slope * x = demand Other Excel Analysis Functions
N Period Moving Average
Let : MAT = The N period moving average at the end of period T AT = Actual observation for period T Then: MAT = (AT + AT-1 + AT-2 + …..+ AT-N+1)/N Characteristics: Need N observations to make a forecast Very inexpensive and easy to understand Gives equal weight to all observations Does not consider observations older than N periods Applicable when no observable trend or seasonality in demand.
Moving Average Example
Saturday Occupancy at a 100-room Hotel
Three-period Moving Average
Saturday
Period
Occupancy
Forecast
Aug.
1 8 15 22 29 Sept. 5 12
1 2 3 4 5 6 7
79 84 83 81 98 100
82 83 87 93
82 83 87 93
Exponential Smoothing
Let : ST = Smoothed value at end of period T AT = Actual observation for period T FT+1 = Forecast for period T+1
Feedback control nature of exponential smoothing New value (ST ) = Old value (ST-1 ) +
? [ observed error ]
ST ? ST-1 ? ? [ AT ? ST ?1 ]
or :
ST ? ? AT ? (1 ? ? ) ST ?1 FT ?1 ? ST
Exponential Smoothing Hotel Example
Saturday Hotel Occupancy (? =0.5) Actual Occupancy At 79 84 83 81 98 100 Smoothed Value St 79.00 81.50 82.25 81.63 89.81 94.91 Forecast Error |At - Ft|
Saturday Aug. 1 8 15 22 29 Sept. 5
Period t 1 2 3 4 5 6
Forecast Ft
79 82 82 82 90
5 1 1 16 10 MAD = 6.6
Forecast Error (Mean Absolute Deviation) = ?lAt – Ftl/n
Exponential Smoothing Implied Weights Given Past Demand
ST ? ?AT ? (1 ? ? ) ST ?1
Substitute for
ST ?1 ? ?AT ? (1 ? ? )[?AT ?1 ? (1 ? ? ) ST ?2 ] ST ? ?AT ? (1 ? ? )[?AT ?1 ? (1 ? ? ) ST ?2 ] ST ? ?AT ? ? (1 ? ? ) AT ?1 ? (1 ? ? ) 2 ST ?2
If continued:
ST ? ?AT ? ? (1 ? ? ) AT ?1 ? ? (1 ? ? ) 2 AT ?2 ?.....?? (1 ? ? ) T ?1 A1 ? (1 ? ? ) T S0
Exponential Smoothing Weight Distribution
0.3
Weight
? ? 0.3
? (1 ? ? ) ? 0.21
? (1 ? ? )2 ? 0147 . ? (1 ? ? )3 ? 0103 . ? (1 ? ? )4 ? 0.072 ? (1 ? ? )5 ? 0.050
0.2 0.1 0 0 1
?
2
3
4
5
Age of Observation (Period Old)
Relationship Between and N 0.1 0.2 19 9 0.3 5.7 0.4 4 0.5 3 0.67 2
?
(exponential smoothing constant) : 0.05 N (periods in moving average) : 39
Saturday Hotel Occupancy
Effect of Alpha (
? =0.1 vs. ?=0.5)
Actual Forecast
(? ? 0.5)
105 100 95 90 85 80 75
0 1 2 3
Occupancy
Forecast
(? ? 0.1)
4
5
Period
6
IT in demand forecasting
?
Demand planning module in a supply chain IT application. IT packages can handle large amount of data and process it quickly with great deal of accuracy. IT tools have built in options for a variety of statistical methods and algorithms. Accuracy of forecast can be quickly tested against historical data to select the most appropriate method.
?
?
?
IT in Demand forecasting
?
IT applicatins allow different methods to be used for different customer segments and different products. What-if analysis can be carried out on IT tools. By capturing transaction data from ERP systems, IT tool can revise estimates on a near real time basis – avoid delayed reaction. IT allows collaborative planning – demand forecast data is fed in to planning module for procurement, production, distribution and inventory planning.
? ?
?
Statistical methods using Time series data:
The First Home Video System
It weighed around 100 lbs by itself, but it was completely transistorized. Extremely high tech for 1963! -The tape speed was 3.75 IPS (inches per second) with recording time of up to 5 hours of black and white video on a single 30 lbs / 12. 5 inch diameter reel of standard 2 inch wide broadcasting video tape. -It Included a 21" color TV, stereo FM tuner, turntable (remember LP's?) and a reel to reel audio tape recorder. A video camera was also included! All Ampex VR-1500 for a modest $30,000! in the 1963 Neiman --Installation included a (free?) visit by Marcus Christmas catalog an Ampex service engineer to set up 900 pounds system.
Home Video War!
?
? ?
Began long before 1974.
Beta vs. VHS Sony introduced Beta Format in 1974. In the first year, Sony sold 30,000 Betamax VCRs in the US. When JVC came out with the VHS format VCR in 1976, the stage was set for the format wars. In 1981, Betamax format VCRs accounted for merely 25% of the entire market and consumers were being warned that the selection for VHS would be "slightly broader.“
?
?
Home Video War!
?
In 1987, Rolling Stone announced that "The battle is over." On January 10, 1988, Sony admitted to plans for a VHS line of VCRs. VHS players commanded 95% of the VCR market.
Laser Disk and DVD formats were under development in 1970’s. One more player! RCA’s Videodisk system. Began field testing in 1975
? ? ?
?
Extensive press coverage in 1976 and 1977
RCA Selecta Videodisk Player
Medium: CED (Capacitance Electronic Discs), ? Housed in 16” housing case -Used a needle to read disc surface ? One sounded: cartridge must be flipped over half way through the movie ? No recording function
?
RCA Selecta Videodisk Player
?
? ? ?
?
?
Demand Forecasting Conducted market research Expected to sell 200,000 players and 2 million disks in 1981 Forecasted that in 10 years the players would be in 30% to 50% of all American households with $7.5 billion in annual sales of players and discs. “Our customer for video discs is clearly the average family, the same broad segment that built the television business to 1980's level of nearly 16 million annual unit sales." (Ack K. Sauter, RCA vice president) “Video disc players will reach a higher sales level in the first year than any other major video product in the history of the industry." Sauter
Why They Did?
?
? ? ? ?
Value Proposition: “Easy to use, depth of software, and affordability” Allies: Zenith, JC Penney, Sears, Sanyo, Toshiba, Hitachi, and Radio Shack 5000 retail outlets $22 Million Advertising Initial Price
» $499.95 for the player » $14.98 to $39.98 for CED disc titles.
What Happened?
?
? ?
?
? ? ?
Sold 100,000 players, half of the forecasted sales, in 1981 When the system hit the market, VCR's were well established Typical consumers thought "Why would I want this VideoDisc player, when for about the same price I can get a VCR that both plays and records.“ RCA's market research didn't take videocassette rental into account at all, and a lot of consumers who earlier would have been willing to purchase movies now preferred to rent them. Cum. Sales in 1984: 500,000 Units Withdrew in 1986 Loss: $580 Million
What We Learn from Videodisk?
?
? ? ?
The importance of Demand Forecasting
Business feasibility assessment Is the market potential big enough? Decision on resource commitment before market launch Pricing
?
Thank you
Exponential Smoothing With Trend Adjustment
St ? ? ( At ) ? (1 ? ? )( St ?1 ? Tt ?1 ) Tt ? ? ( St ? St ?1 ) ? (1 ? ? )Tt ?1 Ft ?1 ? St ? Tt
Commuter Airline Load Factor (? ? 0.5, ? ? 0.3)
Week t 1 2 3 4 5 6 7 8 Actual load factor At 31 40 43 52 49 64 58 68 Smoothed value St 31.00 35.50 39.93 47.10 49.92 58.69 60.88 66.54 Smoothed trend Tt 0.00 1.35 2.27 3.74 3.47 5.06 4.20 4.63 Forecast Forecast error Ft | At - Ft|
31 37 42 51 53 64 65
9 6 10 2 11 6 3 MAD = 6.7
Exponential Smoothing with Seasonal Adjustment
Ferry Passengers taken to a Resort Island (? ? 0.2, ? ? 0.3) Actual Smoothed Index Period t At value St It 2003 January 1 1651 ….. 0.837 February 2 1305 ….. 0.662 March 3 1617 ….. 0.820 April 4 1721 ….. 0.873 May 5 2015 ….. 1.022 June 6 2297 ….. 1.165 July 7 2606 ….. 1.322 August 8 2687 ….. 1.363 September 9 2292 ….. 1.162 October 10 1981 ….. 1.005 November 11 1696 ….. 0.860 December 12 1794 1794.00 0.910 2004 January 13 1806 1866.74 0.876 February 14 1731 2016.35 0.721 March 15 1733 2035.76 0.829
St ? ? ( At / I t ? L ) ? (1 ? ? ) St ?1 Ft ?1 ? ( St )( I t ? L ?1 ) A I t ? ? t ? (1 ? ? ) I t ? L St
Forecast Ft ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. 1236 1653
Error | At - Ft|
495 80
Topics for Discussion
? ?
?
?
?
What characteristics of service organizations make forecast accuracy important? For each of the three forecasting methods, what are the developmental costs and associated cost of forecast error? Suggest independent variables for a regression model to predict the sales volume for a proposed video rental store location. Why is the N-period moving-average still in common use if the simple exponential smoothing model is superior? What changes in ?, ?, ? would you recommend to improve the performance of the trendline seasonal adjustment forecast shown in Figure 11.4?
Demand Forecasting of an Innovative Product/Service
?
What is the challenge of it?
No historical data Time-series methods are inapplicable
?
?
Topics
?
How to forecast sales of a truly innovative product/service How to forecast sales of an early entrant to market How to forecast sales of a pioneer brand after the market entry of a new competitor
?
?
Two Approaches for Sales Forecasting of Innovative Products
?
Survey-Based
Model-Based Diffusion Model (Bass Model) Binary Logit Model
?
?
?
The Objectives of a Diffusion Model
?
To represent a life-cycle sales curve of an innovation-a new product or serviceamong a given set of prospective adopters over time with a small number of parameters.
Assumptions of Diffusion Models
?
Applicable to a new product category not to a new brand.
Each adopter purchases only one unit of the new product No repeat purchases
?
? ? ?
Applicable only to durable goods
Total potential market size is fixed.
The Bass Model
?
Two adopter categories
» Innovators » Imitators
?
Assumption
The new product is first picked up by innovators
?
?
The innovators pass the word (WOM effect) to other membersimitators. The Bass Model (cont’d)
?
S ( t ) = p [ m ? Y ( t )] + qY ( t )[ m ? Y ( t )], m
?
where, S(t) = sales units at time t Y(t) = cumulative sales units at time t Three unknown parameters
» p: adoption rate of innovators » q: adoption rate of imitators » m: potential market size
Run a regression analysis to find a, b, and c.
Problems of a Base Model
?
? ?
?
? ? ? ?
No. of data points required At least 3 sales data points It has been found that the Bass model works well when sales data cover the peak time periods For a truly innovative product, there are no historical sales data. Diffusion process vs Marketing activities One adoption per an adopter One inflection point Application
Application of a Base Model for a Truly Innovative product
?
? ?
There are no historical sales data.
It is better to set the market potential parameter, m,a priori. Products in the same product category tend to have similar Bass coefficients. The innovator coefficients, p, are relatively stable across similar products
?
?
The imitator coefficients, q, tend vary across similar products
Application of a Base Model for a Truly Innovative product
?
Analogical analysis
» Find a set of similar products introduced in the past » Apply the Bass model to these products and get estimates of three Bass parameters » Run a simple linear model
doc_122390458.pptx
forecasting is required, principles of forecasting, factors that influence a forecast, different forecasting methods, time series forecasting.
Forecasting
Forecasting is the art and science of predicting future events Institute of business forecasting
Why Forecast?
Why Forecast?
Forecasts empower people because their use implies that we can modify variables now to alter (or be prepared for) the future. A prediction is an invitation to introduce change into a system. There are several assumptions about forecasting: There is no way to state what the future will be with complete certainty. Regardless of the methods that we use there will always be an element of uncertainty until the forecast horizon has come to pass. There will always be blind spots in forecasts. We cannot, for example, forecast completely new technologies for which there are no existing paradigms. Providing forecasts to policy-makers will help them formulate social policy. The new social policy, in turn, will affect the future, thus changing the accuracy of the forecast.
1.
2.
3.
Why Forecast?
Lead times require that decisions be made in advance of uncertain events. ? Forecasting is an important requirement for all strategic and planning decisions in a supply chain. ? Forecasts of product demand, materials, labor, financing are an important inputs to scheduling, acquiring resources, and determining resource requirements.
?
Demand forecast and Supply Chain Management
Customer’s order is the starting point of initiation of a firm’s business activities, including logistics and supply chain management. If a firm is in a make-to-stock production strategy, SCM is required to plan the level of its activities, such as procurement, transportation, warehousing etc. drugs If a firm in in make-to-order production strategy, SCM is required to plan capacities – machine, labour etc – an aircraft
Demand forecast and Supply Chain Management
In both cases a Supply Chain manager needs a forecast.
Mature markets have a more stable and predictable demand forecast – milk, most food products, items of personal consumption
For unpredictable demand, for example mobile phone models, demand forecast is very difficult – error can cause under or over production and supply.
What is a Forecast Error
?
Demand forecast estimate for a product A is between 30 and 70 units for a period, with a mean of 50 and mean deviation of 20. For a product B the demand forecast estimate is between 40 and 60 units, with the same mean 50 and mean deviation of 10. The 2 firms will be required to organise their logistics differently, if the market demand as per estimates is to be met; with different cost implications. The difference between forecast and actual demand is the forecast error.
?
Principles of Forecasting
?
? ?
?
Forecasts are almost always wrong. Every forecast should include an estimate of the forecast error. The greater the degree of aggregation, the more accurate the forecast. Forecast for total mobile phone demand for India will be more accurate then forecast for sale of Nokia’s N72 model in Pune. Long-term forecasts are usually less accurate than short-term forecasts. Forecasting sale of cold drinks for today and tomorrow is more predictable than for the next week – weather influence, other local factors
Factors that influence a forecast
Demand is influenced by a variety of factors. We must identify each factor that influences demand of the product – » state of economy, » competitors’ activities, » discounts and promotions, » seasonal factors, » substitutes and complimentary products, » and so on.
Factors that influence a forecast
Subjective factors must also be used in correcting demand forecast. A quantitative technique based forecast for umbrellas would not be able to factor the impact of current season’s monsoon activity. Same for woolens, cold drinks and other similar products.
Basic Forecasting Approach
?
?
?
Understand the forecasting objective. What decisions will be made from the forecasts? What parties in the supply chain will be affected by the decision. Integrate demand planning and forecasting. All planning activities within the supply chain that will use the forecast or influence demand should be linked. Collaborative forecast improves response capability to match supply and demand. Identify factors that influence the demand forecast. Is demand growing or declining? Are there relationship (complementary or substitution) between products?
Forecasting Approach (cont.)
?
?
?
Understand and Identify customer segments. Customer demand can be separately forecast for different segments based on service requirements, volume, order frequency, volatility, etc. Demand behaviour for electricity by industries and commercial establishments in festive season differs Determine the appropriate forecasting technique. Typically, using a combination of the different techniques is of the the most effective approach. New products may use judgemental method Establish performance and error measures. Forecasts need to be monitored for their accuracy and timeliness. Eliminate causal effects of a one time promotion.
Forecasting Horizons.
Short Term (0 to 3 months): for inventory management and scheduling. ? Medium Term (3 months to 2 years): for production planning, purchasing, and distribution. ? Long Term (2 years and more): for capacity planning, facility location, and strategic planning.
?
Forecasting Methods
?
?
Qualitative methods are subjective in nature since they rely on human judgment and opinion. Quantitative methods use mathematical or simulation models based on historical demand or relationships between variables.
Selecting a forecast method:
Some Qualitative Methods
?
Jury of Executive Opinion (opinions of a small group of high-level managers is pooled).
Sales Force Composite (aggregation of salespersons estimate of sales in their territory).
?
?Delphi Method (a forecasting group uses a staff to prepare, distribute, collect, and summarize a series of questionnaires and survey results from geographically dispersed respondents, whose judgements are valued).
Some Qualitative Methods
Simulation Methods imitate the consumer choices that give rise to demand to arrive at a forecast. ? Market Research Method (solicit input from customers or potential customers regarding future purchasing plans).
?
Quantitative Forecast Methods
?
?
Time Series Methods use historical data extrapolated into the future. They are best suited for stable environments. Moving averages, exponential smoothing methods, time series decomposition, and Box-Jenkins Methods. Causal Methods assume demand is highly correlated with certain environmental factors (indicators). Correlation methods, regression models, and econometric models.
Time Series Demand Model
?
Observed Demand = Systematic Component + Random Component.
Systematic Component measures the expected value of demand and consists of: ?Level: the current deseasonalized demand. ?Trend: the rate of growth or decline in demand. ?Seasonality: the regular periodic oscillation in demand. Random Component is that part of demand that follows no discernable or predictable pattern.The random component is estimated by the forecast error (forecast – actual demand).
?
?
Time Series Forecasting
?
Static
?Assume estimates of level, trend, and seasonality do not vary as new data is observed.
?
Adaptive
?Update forecast as new data becomes available
Time Series Forecasting
Static ? Adaptive
?
» Moving average » Single exponential smoothing » Trend-adjusted exponential smoothing (Holt’s) » Trend & Seasonal adjusted exponential smoothing (Winter’s)
Static Forecasting (steps)
1. 2. 3.
4.
5.
Determine periodicity of seasonality (even or odd impacts calculation of deseasonalized demand?) Deseasonalize data = Average of consecutive perods demand for all seasons in a cycle Find the equation of the trend line a. Simple linear regression – D = L + Tt - intercept and slope b. Independent variable (period) c. Dependent variable (deseasonalized data) Estimate seasonalized factors a. Per period = observed demand / deseasonalized dem b. Index (Averages) = Average of all period factors Forecast
Find the equation of the line
?
Use simple regression
» Excel: (Tools/Data Analysis/Regression) » Dependent variable
is deseasonalized demand » Independent variable (x) is period t » y= intercept + slope * x = demand Other Excel Analysis FunctionsN Period Moving Average
Let : MAT = The N period moving average at the end of period T AT = Actual observation for period T Then: MAT = (AT + AT-1 + AT-2 + …..+ AT-N+1)/N Characteristics: Need N observations to make a forecast Very inexpensive and easy to understand Gives equal weight to all observations Does not consider observations older than N periods Applicable when no observable trend or seasonality in demand.
Moving Average Example
Saturday Occupancy at a 100-room Hotel
Three-period Moving Average
Saturday
Period
Occupancy
Forecast
Aug.
1 8 15 22 29 Sept. 5 12
1 2 3 4 5 6 7
79 84 83 81 98 100
82 83 87 93
82 83 87 93
Exponential Smoothing
Let : ST = Smoothed value at end of period T AT = Actual observation for period T FT+1 = Forecast for period T+1
Feedback control nature of exponential smoothing New value (ST ) = Old value (ST-1 ) +
? [ observed error ]
ST ? ST-1 ? ? [ AT ? ST ?1 ]
or :
ST ? ? AT ? (1 ? ? ) ST ?1 FT ?1 ? ST
Exponential Smoothing Hotel Example
Saturday Hotel Occupancy (? =0.5) Actual Occupancy At 79 84 83 81 98 100 Smoothed Value St 79.00 81.50 82.25 81.63 89.81 94.91 Forecast Error |At - Ft|
Saturday Aug. 1 8 15 22 29 Sept. 5
Period t 1 2 3 4 5 6
Forecast Ft
79 82 82 82 90
5 1 1 16 10 MAD = 6.6
Forecast Error (Mean Absolute Deviation) = ?lAt – Ftl/n
Exponential Smoothing Implied Weights Given Past Demand
ST ? ?AT ? (1 ? ? ) ST ?1
Substitute for
ST ?1 ? ?AT ? (1 ? ? )[?AT ?1 ? (1 ? ? ) ST ?2 ] ST ? ?AT ? (1 ? ? )[?AT ?1 ? (1 ? ? ) ST ?2 ] ST ? ?AT ? ? (1 ? ? ) AT ?1 ? (1 ? ? ) 2 ST ?2
If continued:
ST ? ?AT ? ? (1 ? ? ) AT ?1 ? ? (1 ? ? ) 2 AT ?2 ?.....?? (1 ? ? ) T ?1 A1 ? (1 ? ? ) T S0
Exponential Smoothing Weight Distribution
0.3
Weight
? ? 0.3
? (1 ? ? ) ? 0.21
? (1 ? ? )2 ? 0147 . ? (1 ? ? )3 ? 0103 . ? (1 ? ? )4 ? 0.072 ? (1 ? ? )5 ? 0.050
0.2 0.1 0 0 1
?
2
3
4
5
Age of Observation (Period Old)
Relationship Between and N 0.1 0.2 19 9 0.3 5.7 0.4 4 0.5 3 0.67 2
?
(exponential smoothing constant) : 0.05 N (periods in moving average) : 39
Saturday Hotel Occupancy
Effect of Alpha (
? =0.1 vs. ?=0.5)
Actual Forecast
(? ? 0.5)
105 100 95 90 85 80 75
0 1 2 3
Occupancy
Forecast
(? ? 0.1)
4
5
Period
6
IT in demand forecasting
?
Demand planning module in a supply chain IT application. IT packages can handle large amount of data and process it quickly with great deal of accuracy. IT tools have built in options for a variety of statistical methods and algorithms. Accuracy of forecast can be quickly tested against historical data to select the most appropriate method.
?
?
?
IT in Demand forecasting
?
IT applicatins allow different methods to be used for different customer segments and different products. What-if analysis can be carried out on IT tools. By capturing transaction data from ERP systems, IT tool can revise estimates on a near real time basis – avoid delayed reaction. IT allows collaborative planning – demand forecast data is fed in to planning module for procurement, production, distribution and inventory planning.
? ?
?
Statistical methods using Time series data:
The First Home Video System
It weighed around 100 lbs by itself, but it was completely transistorized. Extremely high tech for 1963! -The tape speed was 3.75 IPS (inches per second) with recording time of up to 5 hours of black and white video on a single 30 lbs / 12. 5 inch diameter reel of standard 2 inch wide broadcasting video tape. -It Included a 21" color TV, stereo FM tuner, turntable (remember LP's?) and a reel to reel audio tape recorder. A video camera was also included! All Ampex VR-1500 for a modest $30,000! in the 1963 Neiman --Installation included a (free?) visit by Marcus Christmas catalog an Ampex service engineer to set up 900 pounds system.
Home Video War!
?
? ?
Began long before 1974.
Beta vs. VHS Sony introduced Beta Format in 1974. In the first year, Sony sold 30,000 Betamax VCRs in the US. When JVC came out with the VHS format VCR in 1976, the stage was set for the format wars. In 1981, Betamax format VCRs accounted for merely 25% of the entire market and consumers were being warned that the selection for VHS would be "slightly broader.“
?
?
Home Video War!
?
In 1987, Rolling Stone announced that "The battle is over." On January 10, 1988, Sony admitted to plans for a VHS line of VCRs. VHS players commanded 95% of the VCR market.
Laser Disk and DVD formats were under development in 1970’s. One more player! RCA’s Videodisk system. Began field testing in 1975
? ? ?
?
Extensive press coverage in 1976 and 1977
RCA Selecta Videodisk Player
Medium: CED (Capacitance Electronic Discs), ? Housed in 16” housing case -Used a needle to read disc surface ? One sounded: cartridge must be flipped over half way through the movie ? No recording function
?
RCA Selecta Videodisk Player
?
? ? ?
?
?
Demand Forecasting Conducted market research Expected to sell 200,000 players and 2 million disks in 1981 Forecasted that in 10 years the players would be in 30% to 50% of all American households with $7.5 billion in annual sales of players and discs. “Our customer for video discs is clearly the average family, the same broad segment that built the television business to 1980's level of nearly 16 million annual unit sales." (Ack K. Sauter, RCA vice president) “Video disc players will reach a higher sales level in the first year than any other major video product in the history of the industry." Sauter
Why They Did?
?
? ? ? ?
Value Proposition: “Easy to use, depth of software, and affordability” Allies: Zenith, JC Penney, Sears, Sanyo, Toshiba, Hitachi, and Radio Shack 5000 retail outlets $22 Million Advertising Initial Price
» $499.95 for the player » $14.98 to $39.98 for CED disc titles.
What Happened?
?
? ?
?
? ? ?
Sold 100,000 players, half of the forecasted sales, in 1981 When the system hit the market, VCR's were well established Typical consumers thought "Why would I want this VideoDisc player, when for about the same price I can get a VCR that both plays and records.“ RCA's market research didn't take videocassette rental into account at all, and a lot of consumers who earlier would have been willing to purchase movies now preferred to rent them. Cum. Sales in 1984: 500,000 Units Withdrew in 1986 Loss: $580 Million
What We Learn from Videodisk?
?
? ? ?
The importance of Demand Forecasting
Business feasibility assessment Is the market potential big enough? Decision on resource commitment before market launch Pricing
?
Thank you
Exponential Smoothing With Trend Adjustment
St ? ? ( At ) ? (1 ? ? )( St ?1 ? Tt ?1 ) Tt ? ? ( St ? St ?1 ) ? (1 ? ? )Tt ?1 Ft ?1 ? St ? Tt
Commuter Airline Load Factor (? ? 0.5, ? ? 0.3)
Week t 1 2 3 4 5 6 7 8 Actual load factor At 31 40 43 52 49 64 58 68 Smoothed value St 31.00 35.50 39.93 47.10 49.92 58.69 60.88 66.54 Smoothed trend Tt 0.00 1.35 2.27 3.74 3.47 5.06 4.20 4.63 Forecast Forecast error Ft | At - Ft|
31 37 42 51 53 64 65
9 6 10 2 11 6 3 MAD = 6.7
Exponential Smoothing with Seasonal Adjustment
Ferry Passengers taken to a Resort Island (? ? 0.2, ? ? 0.3) Actual Smoothed Index Period t At value St It 2003 January 1 1651 ….. 0.837 February 2 1305 ….. 0.662 March 3 1617 ….. 0.820 April 4 1721 ….. 0.873 May 5 2015 ….. 1.022 June 6 2297 ….. 1.165 July 7 2606 ….. 1.322 August 8 2687 ….. 1.363 September 9 2292 ….. 1.162 October 10 1981 ….. 1.005 November 11 1696 ….. 0.860 December 12 1794 1794.00 0.910 2004 January 13 1806 1866.74 0.876 February 14 1731 2016.35 0.721 March 15 1733 2035.76 0.829
St ? ? ( At / I t ? L ) ? (1 ? ? ) St ?1 Ft ?1 ? ( St )( I t ? L ?1 ) A I t ? ? t ? (1 ? ? ) I t ? L St
Forecast Ft ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. 1236 1653
Error | At - Ft|
495 80
Topics for Discussion
? ?
?
?
?
What characteristics of service organizations make forecast accuracy important? For each of the three forecasting methods, what are the developmental costs and associated cost of forecast error? Suggest independent variables for a regression model to predict the sales volume for a proposed video rental store location. Why is the N-period moving-average still in common use if the simple exponential smoothing model is superior? What changes in ?, ?, ? would you recommend to improve the performance of the trendline seasonal adjustment forecast shown in Figure 11.4?
Demand Forecasting of an Innovative Product/Service
?
What is the challenge of it?
No historical data Time-series methods are inapplicable
?
?
Topics
?
How to forecast sales of a truly innovative product/service How to forecast sales of an early entrant to market How to forecast sales of a pioneer brand after the market entry of a new competitor
?
?
Two Approaches for Sales Forecasting of Innovative Products
?
Survey-Based
Model-Based Diffusion Model (Bass Model) Binary Logit Model
?
?
?
The Objectives of a Diffusion Model
?
To represent a life-cycle sales curve of an innovation-a new product or serviceamong a given set of prospective adopters over time with a small number of parameters.
Assumptions of Diffusion Models
?
Applicable to a new product category not to a new brand.
Each adopter purchases only one unit of the new product No repeat purchases
?
? ? ?
Applicable only to durable goods
Total potential market size is fixed.
The Bass Model
?
Two adopter categories
» Innovators » Imitators
?
Assumption
The new product is first picked up by innovators
?
?
The innovators pass the word (WOM effect) to other membersimitators. The Bass Model (cont’d)
?
S ( t ) = p [ m ? Y ( t )] + qY ( t )[ m ? Y ( t )], m
?
where, S(t) = sales units at time t Y(t) = cumulative sales units at time t Three unknown parameters
» p: adoption rate of innovators » q: adoption rate of imitators » m: potential market size
Run a regression analysis to find a, b, and c.
Problems of a Base Model
?
? ?
?
? ? ? ?
No. of data points required At least 3 sales data points It has been found that the Bass model works well when sales data cover the peak time periods For a truly innovative product, there are no historical sales data. Diffusion process vs Marketing activities One adoption per an adopter One inflection point Application
Application of a Base Model for a Truly Innovative product
?
? ?
There are no historical sales data.
It is better to set the market potential parameter, m,a priori. Products in the same product category tend to have similar Bass coefficients. The innovator coefficients, p, are relatively stable across similar products
?
?
The imitator coefficients, q, tend vary across similar products
Application of a Base Model for a Truly Innovative product
?
Analogical analysis
» Find a set of similar products introduced in the past » Apply the Bass model to these products and get estimates of three Bass parameters » Run a simple linear model
doc_122390458.pptx