COMMERCE 2OC3 Study Guide - Midterm Guide: Mean Squared Error, Dependent And Independent Variables, Average Absolute Deviation
Forecasting is the science of prediction.
Type of Forecast in Business
Example
Economic
Interest rates, inflation, macroeconomic supply and demand
Technological
Moore's Law, the theory that computer power doubles every 12-18 months
Type of Forecast in Operations
Example
Demand
Projections of demand for company products/services
HRM - E.g. Disney World cast members are scheduled in 15 minute intervals based on demand forecasts
•
Capacity management - waste or loss of capacity occurs when capacity > demand, and demand > capacity - E.g. Uber dynamic
pricing changes depending on demand forecasts
•
Supply chain management - E.g. Amazon fulfillment centres/warehouses concentrates in east/west coast USA, India, East Asia
•
Purpose of Forecasting in Operations
Forecasts are always wrong and have some degree of error
1.
The longer the horizon, the larger the error such that forecasts should be updated periodically
2.
Principles of Forecasting
Use human intelligence compared to quantitative forecasting's artificial intelligence
•
Qualitative Forecasting
Methods
Description
Jury of executive opinion
Pooled opinion of group of experts in field - e.g. "220 scientists used to predict future trends in…"
Delphi method
Structure communicating method developed during Cold War by RAND Corporation
"How many bombs need to be delivered the by the USSR to reduce the US munitions output to
one quarter?"
Round 1: estimates by experts are extremely varied
▪
Round 2: other experts' opinion/analysis incorporated
▪
Round 5: estimate by experts are more uniform
▪
Panel of 7 anonymous experts in five rounds:
Sales force composite
Estimation by sales agents involving questions about which products are actually selling
Consumer market survey
Surveying potential customers' future purchasing plans
Quantitative Forecasting Methods
Time series models
Sequence of evenly paced data points
•
Trend describes gradual upward or downward movement of data over time
•
Seasonality indicates regular vertical patterns oscillating along the trendline - there are
multiple within a given year with roughly the same duration
•
Random variation describes the presence of erratic fluctuation with no discernable pattern
•
Cycles describe vertical patterns occurring every few years - e.g. economic cycle averaging
5.5 years in the USA, with wide variation ranging from 18 months to 10 years - indefinite
length
•
Time window moves, with greater data points available to analyze, but most outliers that skew the average
•
Naïve approach
= forecast of period
= actual demand of period
Moving averages
Weight depends on importance of previous periods - use if n periods are not equally important, e.g. winter apparel sales does
not have a huge impact on summer apparel sales
•
Weighted moving averages
Forecasting (52+35+38)
February 8, 2018
2:32 PM
Operations Management Page 1