MODULE 2: Demand Forecasting FORECASTING

 Art and science of predicting a future

event
TOPICS:  Underlying basis of all business
decisions
-WHAT IS FORECASTING? - Production
- Inventory
 Forecasting Time Horizon
- Personnel
 Influence of Product Life Cycle
- Facilities
 Types of Forecast
FORECASTING TIME HORIZONS
-STRATEGIC IMPORTANCE OF FORECASTING
 SHORT-RANGE FORECAST
-FORECASTING APPROACHES
- Up to 1 year, generally less than 3
 Qualitative Forecasting Models months
 Quantitative Forecasting Models - Purchasing, job scheduling workforce
levels. Job assignments, production
levels
-TIME SERIES FORECASTING - tend to be more accurate than longer-
term forecast
 Decomposition of a Time Series
 Naïve Approach
 MEDIUM-RANGE FORECAST
 Moving Averages
- 3 months to 3 years
 Exponential Smoothing
- Sales and production, planning,
 Exponential Smoothing with Trend
budgeting
Adjustment
 Trend Projections
 LONG-RANGE FORECAST
 Seasonal Variations in Data
- 3+ years
 Cyclical Variations in Data
- New products planning facility
location, research and development
-ASSOCIATIVE FORECASTING METHODS: - deal with more comprehensive issues
REGRESSION AND CORRELATION ANALYSIS
INFLUENCE OF PRODUCT LIFE CYCLE
 Using Regression Analysis for
 INTRODUCTION-GROWTH-MATURITY-
Forecasting
DECLINE
 Standard Error Estimate
 Introduction and growth require longer
 Correlation Coefficients for Regression
forecasts than maturity and decline
Lines
 As product passes through life cycle,
 Multiple Regression Analysis
forecasts are useful in projecting;
staffing levels, inventory levels, factory
-MONITORING AND CONTROLLING FORECAST capacity

 Adaptive Smoothing
 Focus Forecasting

- FORECASTING IN THE SERVICE SECTOR

TYPES OF FORECAST  Jury of Executive Opinion- Pool
opinions of high-level experts,
 ECONOMIC FORECAST- addresses sometimes augment by statistical
business cycle- inflation rate, money
models
supply, housing starts, etc.  Delphi Model- Panel of experts, queried
iteratively , 3 types of Participants:
 TECHNOLOGICAL FORECASTS- predicts Decision Makers, Staff, Respondents
rate of technological progress, impacts  Sales force composite- Estimates from
development of new products
individual salespersons are reviewed for
reasonableness, then aggregates
 DEMAND FORECASTS- predicts sales of  Consumer Market Survey- Ask the
existing product and services customer
STRATEGIC IMPORTANCE OF FORECASTING
 QUANTITATIVE METHODS
 HUMAN RESOURCES- Hiring, training,
laying off workers - Used when situation is stable and
 CAPACITY- Capacity shortages can historical data exist: Existing products,
result in undependable delivery, loss of Current technology
customers, loss of market share
- Involves mathematical techniques:
 SUPPLY CHAIN MANAGEMENT- Good forecasting sales of color television
supplier relations and price advantages
 OVERVIEW OF QUANTITATIVE
SEVEN STEPS IN FORECASTING METHODS
1. Determine the use of the forecast
2. Select the items to be forecasted  TIME SERIES MODELS:
3. Determine the time horizon of the o Naïve Approach
forecast o Moving Averages
4. Select the forecasting model(s) o Exponential Smoothing
5. Gather the data  ASSOCIATIVE MODEL
6. Make the forecast o Trend Projection
7. Validate and implement results o Linear Regression

NOTE: REALITIES TIME SERIES FORECASTING

 Forecasts are seldom perfect  Set of evenly spaced numerical data

 Most techniques assume an underlying  Forecast based only on past values, no
stability in the system other variables important
 Product family aggregated forecasts
TIME SERIES COMPONENTS
are more accurate than individual
product forecasts  TREND COMPONENT- persistent,
changes due to population,tech etc.,
FORECASTING APPROACHES
typically several years duration
 QUALITATIVE METHODS  SEASONAL COMPONENT- Regular
pattern of up and down fluctuations,
- Used when situation is vague and little due to weather, occurs within a single
data exist: New products, New technology year
- Involves intuition, experience: Forecasting  CYCLICAL COMPONENT- Repeating up
sales on internet and down movements, affected by
business cycle, political etc., multiple
 OVERVIEW OF QUALITATIVE METHODS years duration
  RANDOM COMPONENT- Erratic, POTENTIAL PROBLEMS WITH MOVING
unsystematic, residual fluctuations, due AVERAGE:
to random or unforeseen events, short
 Increasing n smooth the forecast but
duration and non-repeating
makes it less sensitive to changes
 Do not forecast trends well
 NAÏVE APPROACH
 Require extensive historical data
- assumes demand in next period is
the same as demand in most recent
 EXPONENTIAL SMOOTHING
period
 Form of weighted moving average
- sometimes cost effective and
o Weights decline exponentially
efficient, can be a good starting
o Most recent data weighted
point
most
 MOVING AVERAGE METHOD  Requires smoothing constant ()
 MA is a series of arithmetic means o Ranges from 0 to 1
 Used if little or no trend o Subjectively chosen
 Used often for smoothing  Involves little record keeping of
 Provides overall impression of data over past data
time
𝒅𝒆𝒎𝒂𝒏𝒅 𝒊𝒏 𝒑𝒓𝒆𝒗𝒊𝒐𝒖𝒔 𝒏 𝒑𝒆𝒓𝒊𝒐𝒅𝒔
𝑴𝒐𝒗𝒊𝒏𝒈 𝑨𝒗𝒆𝒓𝒂𝒈𝒆 =
𝒏
EXAMPLE:

 WEIGHTED MOVING AVERAGE

 Used when some trend might be
present
 Older data usually less important
IMPACT OF DIFFERENT ALPHA
 Weights based on experience and
intuition  Chose high values of  when
underlying average is likely to
change
EXAMPLE  Choose low values of  when
underlying average is stable

CHOOSING ALPHA

- We generally do this by selecting the

model that gives us the lowest forecast
error.

Forecast error = Actual demand -

Forecast value

= At - Ft
  COMMON MEASURES OF ERROR  EXPONENTIAL SMOOTHING WITH
TREND ADJUSTMENT


Example:

Example:

Example: