Forcasting - DR. Gavida forecasting PDF

Preview

Summary

DR. Gavida forecasting...

Description

Intro Into Forecasting (video) Demand Forecasting 1. Definitions a. Forecasting (basically a guess) i. Have errors ii. Calculations are not forecasts iii. Plans are not forecasts b. Demand forecasting and SCM i. Consumer demand drives activities of the entire supply chain and demand from customers will influence … ii. A plan is not a forecast c. Quantitative and qualitative forecasting i. Delphi method 1. a process used to arrive at a group opinion or decision by surveying a panel of experts. Experts respond to several rounds of questionnaires, and the responses are aggregated and shared with the group after each round 2. Forecast time horizons a. When there is uncertainty in demand forecasting companies re taking risks because if they are planning to do something thats not going to happen and no demand for they are going to lose money so making wrong forecast can lead to losing a lot of money b. Forecasting is what allows comps to match supply and demand i. Forecast correct amt of demand then supply will match demand and there is efficient production, and vice versa 3. Time horizons and forecasting models a. Short term forecasting i. Time= daily/weekly ii. data= automatic collection of common ops nd sales metrics, automatic execution iii. cost= low , automated forecasting iv. use= master production scheduling, purchasing v. Aggregation= single products vi. Common forecasting methods= naive forecasts, moving avgs, exponential smoothing b. Intermediate term forecasting i. Time= monthly/annual ii. data = automatic data collection sometimes combined with external economic and competitor data

iii. cost= medium, some data collections and analysis iv. use= sales and ops planning v. aggregation= product groups or families vi. Common forecasting methods= trend analysis, qualitative methods c. Long term forecasting i. time= more than 2 years, up to several decades for infrastructure decisions ii. data= long term trends and macro indicators used in custom forecasts iii. cost= high, extensive data collection and research iv. use= strategic decision making, infrastructures, new products, market entry v. aggregation= total sales of business units or major subdivisions vi. Common forecasting methods= qualitative methods, trend analysis with seasonality casual methods such as multiple regression, econometrics 4. Forecasting models a. Regression i. Cost is the indicator (x variable) ii. Effect is the sales (y variable) iii. Consumer confidence allows us to forecast sales into the future 1. Focus on leading indicators which allow us to forecast future where lagging indicators confirm what happened in the past iv. y=a +bx b. Constant model i. Rarely any change ii. Moving avg iii. Weighted moving avg iv. Exponential smoothing v. Constant over time c. Trend model i. Time series: trend ii. Tendency of the data to increase or decrease with time iii. Time=cost; sales= effect d. Seasonal model i. Time series: seasonality e. Seasonal trend model i. Time series: seasonality and trend

Long term forecasting models 1. Forecasting models

a. Causal model ( theres a cause and effect) i. Regression ii. Independent variable = indicator and dependent variable= trying to forecast (demand or sales) b. Constant models i. Trend model ii. Long term regression model iii. Very similar to regression only different is independent variable is time( the cause of the change) c. Seasonal model i. Seasonal pattern that takes place over time, repeating d. Seasonal trend model i. Advise to estimate seasonality and then be able to forecast future trend by extrapolating it and seasonalized forecast trend e. Long term forecasting models i. Focusing on: 1. Causal analysis a. Working an example with leading indicators to forecast demand (consumer confidence = leading indicator) i. Will be a lead time where indicator doesnt forecast current period , learn how to align variable 2. Time series decomposition a. Use a model with seasonality and trend 2. Causal analysis a. Basics i. Independent variable (x) ii. Dependent variable (y) iii. Model or function that describes (y=f(x)) b. Simple linear regression i. y=f(x) ii. Problem: consumer confidence

Short term forecastingALL ONLY GOOD FOR FORECASTING NEXT PERIOD OF DEMAND 1. Naive a. Forecast for next period for period t +1 is = to the actual full period t b. Random walks

i.

2.

3.

4.

5.

Situation when really no idea what will happen, completely random like stock prices of tomorrow c. Constant variables d. Benchmark for perf Simple moving averages (DRIVING XS PROVPEM) a. N= order or moving average b. Balance between responsiveness and stability c. As you increase order of the simple moving average you make it more stable and as decrease it becomes more responsive to last few observations d. All weights are the same Weighted moving averages a. More recent observations have higher weights b. We can manage responsiveness and stability more precisely than MA c. Can control how many and actually distribute weights individually First order Exponential smoothing a. Makes correction of previous error b. alpha= smoothing constant c. Larger values are more responsive to Winters’ forecasting method a. Parameters i. Alpha for basic value ii. Beta for slope iii. Gamma for seasonality

Accuracy of forecasting models 1. Ex-post forecasts a. Accuracy b. Outliers c. Parameters...