Forecasting and Demand Planning

1
Forecasting and Demand Planning

• Chapter 11
• Operations Management Goods, Services, and
  Value Chains, by D. A. Collier and J. R. Evans

2
Why pay the cost of forecasting?

• Good forecasting leads to good decisions.
• Operations Dysfunctions Due to Bad Forecasts
• Poor Staffing of Call Centers
• Inadequate Inventories
• Parts shortages
• Excess Capacity
• Service Delivery Delays
• Production equipment idling

3
The Need for Forecasts in a Value Chain
4
Discussion Questions

• Think of a pizza delivery franchise located near
  a college campus.
• What factors that influence demand do you think
  should be included in trying to forecast demand
  for pizzas?
• How might these factors differ for a franchise
  located in a suburban residential area?

5
Collaborative Demand Planning

• By sharing information across the value chain
  about customer order status, customer and
  supplier delivery schedules, backorders, and
  inventory status, companies in the value chain
  reduce their need for forecasts and also improve
  the accuracy of the forecasts they have to make.

6
Definitions

• Forecast Planning Horizon Length
• Long Term (1-10 years) difficult to predict
• Medium Term (3-12 months) easier to predict
• Short Term ( lt 3 months) relatively high
  forecast accuracy
• Time series are formed by sequential observations
  of the same variable
• Number of observations in regular series depends
  on span between observations or time bucket

7
Data Patterns in Times Series

• Trend
• Up, Down, no Trend, Linear, Non Linear
• Seasonal
• Yearly, Weekly, Monthly
• Cyclical
• Cycles without fixed period (as Seasonal)
• Random variation
• Irregular (one-time) variation.

8
Linear Trend
9
Trend Patterns
10
Seasonal Pattern
11
Cyclical Pattern
12
Forecast Errors and Accuracy

• Forecast error is the difference between observed
  value of time series and forecast, or At-Ft.

13
Types of Forecasting Approaches
14
Single Moving Average
15
Single Exponential Smoothing
16
Advanced Forecasting Models

• Double moving averageused for time series with a
  linear trend
• Double exponential smoothingused for time series
  with a linear trend
• Seasonal additiveused for time series with
  seasonality that is relatively stable over time
• Seasonal multiplicativeused for time series with
  seasonality that is increasing or decreasing in
  magnitude over time
• HoltWinters additiveused for time series with
  both a linear trend and seasonality that is
  relatively stable over time
• HoltWinters multiplicativeused for time series
  with both a linear trend and seasonality that is
  increasing or decreasing in magnitude over time

17
Regression Trendline
18
Causal Forecasting Model

• Time is one variable
• Another variable is /gallon (causal)
• Use Multiple Regression to forecast week 11

19
Causal Forecasting Model