Forecasting 1 Linkages How much we are going to sell is obviously important to marketing Forecasts help us to plan investments - or to determine if an investment is a good idea Forecasts tell us if we will have to hire new people and or train our existing people in new skills Technological forecasts might indicate the need to change our MIS function
Forecasting 2 Forecasts as part of planning How much demand we are going to have leads to a number of other questions –large demand for standard products: line flow –demand for custom products: jumbled flow –demand leads to capacity –demand indicates when we schedule work –etc. In other words a forecast is one of the first things we need when planning – for the long term and the short term.
Forecasting 3 Why do forecasts matter ? People: If we do a bad job forecasting demand we may not have the right number (or type) of people on hand. Capacity: If we under forecast we will not be able to make enough stuff (lost sales) over forecasting will result in expensive wasted capacity. Supply chain: Our suppliers are also dependent on our forecasts: –What if we have them build stuff based on an erroneously high forecast?
Forecasting 4 Characteristics of forecasts Short term: Less than a year –quantitative –can be very accurate –dis-aggregated Long term: More than a year –often very qualitative –much harder to be accurate –generally aggregated
Forecasting 5 Types of forecasts Economic: What is happening in the world, country, state, and locality. Aggregated across companies and usually industries. –ISM index –The federal reserve Technological: changes in technology that may change products and / or processes –BW survey of research labs Demand: Sales of our company’s products - often driven (partially) by economic and technological.
Forecasting 6 Quantitative verses Qualitative When numbers do not exist and or are inaccurate we can use qualitative methods (long term forecasts especially) –delphi methods –market research –the “gut” Most people want to use numbers –why? –is this always best? See readings on methods people do choose- and what would be best.
Forecasting 7 Forecasting demand There are 5 components of demand: –Average demand – not in book –Trends –Cyclicality –Seasonality –Random factors What should we be able to forecast ?
Forecasting 8 Trend Sales of Dallas Cowboys Paraphernalia Volume Year projected
Forecasting 9 Seasonality Beverage sales at the 6 pac shop MONMON TUSTUS WENWEN TURTUR FRIFRI SATSAT SUNSUN MONMON
Forecasting 10 Seasonality 2 Umbrella sales SummerSummer SummerSummer FallFall FallFall WinterWinter WinterWinter SpringSpring SpringSpring
Forecasting 11 Cyclicality The business cycle –Where are we in the business cycle? –to forecast the end of a period of growth what signs would you look for? What do you think Greenspan looks for?
Forecasting 12 Determining the quality of a forecast Forecast error = demand - forecast negative errors indicate ? Mean Absolute Deviation (MAD) Mean percentage deviation (MAPE)
Forecasting 13 Determining the quality of a forecast 2 Why don’t we use the average deviation? What does the MAPE tell us that the MAD does not ? –can we compare the MADS for two different products ? –can we use MAD to compare the same forecasting method in a variety of situations ? BiasWe also want to examine Bias
Forecasting 14 MAD / MAPE example
Forecasting 15 A quick aside The forecasting tools we are going to use are generally basic and fairly simple. –See the articles I placed on the web- this is what people use –Regression is “to fancy” for many managers Our goal is to find the method that best fits our pattern of demand- no one right tool
Forecasting 16 Actual forecasting tools The simplest method: the naive forecast –this period’s demand = last period’s demand –when is this acceptable ? Time series methods: future demand is predicted from past (historical) demand. –moving averages simple and weighted –exponential smoothing
Forecasting 17 Moving averages A simple tool to predict demand when it is safe to assume that over time demand is fairly stable (change is slow). A 3 period moving average: A five period moving average:
Forecasting 18 Moving average example
Forecasting 19 Weighted moving averages Moving averages work fine when the world is fairly stable - but what if our world is changing ? Weighted moving averages (WMA) - place more weight on recent events (why). WMA = (Σ (weight period n) (demand in period n)) / Σ weights Determining weights is an art - generally do not weight most recent period more than 50%.
Forecasting 20 AWMA example: Weights 5,3,2
Forecasting 21 Exponential smoothing Exponential smoothing is a very popular (and simple) form of the weighted moving average. Basic form: What happens as the smoothing constant increases ?
Forecasting 22 Exponential smoothing: examples smoothing constant =.2 perioddemandforecast (4) = (2.2) = smoothing constant =.5 perioddemandforecast (4) = (1) = 23.5
Forecasting 23 Seasonality Because seasonality is a pattern we can predict it using indices. For example: yearly demand = 800 units indices: spring =.85 summer = 1.46 fall =.76 winter =.93 –F spring = 200 *.85 = 170 –F summer = 200 * 1.46 = 292 –f fall = 200 *.76 = 152 –f winter = 200 *.93 = 186
Forecasting 24 Determining Indices
Forecasting 25 More indices stuff The sum of the indices should = the number of seasons. Formula for the index : average demand specific season average demand all seasons
Forecasting 26 Regression Models The basic regression model –F = constant + b 1 X 1 + b 2 X 2 – b 1 is a constant – X 1 is an independent variable –You can of course use only 1 independent variable in your model- or more than 2 (sometimes many more than 2)
Forecasting 27 Some obvious uses of regression We can use regression to forecast when we have a trend in the data. –If the trend is the major source of change in the data might be able to use a simple regression model where time is our only independent variable Ft = constant + b (time period) We might also make the season an independent variable We can obviously include just about anything else in the model that makes sense demand for ice cream - might include temperature demand for MBA classes - unemployment rate
Forecasting 28 Other issues Double exponential smoothing –Fancy way to try and cope with trends – works well when there is a one time change – not as well when the trend is always going up / down Focused forecasting –Economist
Forecasting 29 Book problems you should be able to do 4.2, 4.4, 4.6, 4.8, 4.10, 4.26, 4.28