Chapter Three

Forecasting
Learning outcome
 Define Forecasting
Importance of Forecasting
Types of Forecasting
Techniques of Forecasting
 Qualitative
 Quantitative
1
 Introduction to Forecasting
A statement about the future value of a variable of
interest such as demand.
Forecasting is a tool used for predicting future
demand based on past demand information.
 Forecasts affect decisions and activities throughout
an organization.
Accounting, finance
Human resources
Marketing
Management information system (MIS),
Operations, Product / service design.
2
 Importance of Forecasting
 More specifically, here are who and why they need to
forecast:
 Marketing managers: use sales forecasts to determine
optimal sales force allocations set sales goals, and plan
promotions and advertising.
 Planning for capital investments: predictions about
future economic activity are required so that returns or
cash inflows accruing from the investment may be
estimated.
 The personnel department requires a number of
forecasts in planning for human resources.
 Managers of nonprofit institutions and public
administrators also must make forecasts for budgeting
purposes. 3
Universities: forecast student enrollments, cost of
operations, and, in many cases, the funds to be provided
by tuition and by government appropriations.
The bank has to forecast: Demands of various loans
and deposits Money and credit conditions so that it can
determine the cost of money it lends.
 Demand is not the only variable of interest to
forecasters.
 Manufacturers: also forecast worker absenteeism,
machine availability, material costs, transportation
and production lead times, etc.
 Besides demand, service providers are also interested
in forecasts of population, of other demographic
variables, of weather, etc.

4
Accounting Cost/profit estimates

Finance Cash flow and funding

Human Resources Hiring/recruiting/training

Marketing Pricing, promotion, strategy

MIS IT/IS systems, services

Operations Schedules, MRP, workloads

Product/service design New products and services

5
 Forecasting Range
Short-range forecasts: typically
encompass the immediate future and
are concerned with the daily
operations of a business firm, such as
daily demand or resource
requirements. A short-range forecast
rarely goes beyond a couple months
into the future.

6
 A medium-range forecast: typically encompasses
anywhere from 1 or 2 months to 1 year.
 A forecast of this length is generally more closely related to
a yearly production plan and will reflect such items as peaks
and valleys in demand and the necessity to secure additional
resources for the upcoming year.
A long-range forecast: typically encompasses a period
longer than 1 or 2 years. Long-range forecasts are related to
management's attempt to plan new products for changing
markets, build new facilities, or secure long-term financing.
In general, the further into the future one seeks to predict,
the more difficult forecasting becomes.
7
Short-range forecast Quantitative
methods
Usually < 3 months
 Job scheduling, worker assignments Detailed
use of
Medium-range forecast system
3 months to 2 years
 Sales/production planning
Long-range forecast
> 2 years
Design
 New product planning of system
Qualitative
Methods
 Principles of Forecasting
Many types of forecasting models that differ in
complexity, amount of data & way they generate
forecasts:-

Forecasts rarely perfect because of randomness.

Forecasts more accurate for groups vs. individuals.

Forecast accuracy decreases as time horizon

increases.

9
Steps of Forecasting
1. Decide what needs to be forecasted.
Level of detail, units of analysis & time horizon

required.
2. Evaluate and analyze appropriate data.
Identify needed data & whether it’s available.

3. Select and test the forecasting model.

Cost, ease of use & accuracy.

4. Generate the forecast.

5. Monitor forecast accuracy over time.

10
Forecasting Techniques
Forecasting practice is based on a mix of qualitative
and quantitative methods.

When planning occurs for innovative products, little

demand data are available for the product of interest
and the degree to which like product demand data are
similarly unknown.

Thus a large amount of judgment is needed by experts

who can use their industry expertise to predict demand.
11
Commodity-like products that are sold every day, on the
other hand, are much more suitable for quantitative
models and need very little judgment to forecast
demand.
Still, when knowledge of certain events leads one to
believe that future demand might not track historical
trends, some judgment may be warranted to make
adjustments in the models which use past data.

In this case, a heavy reliance on past data with

adjustments based on expert judgment should be the
method used for forecasting.
12
 Qualitative methods – judgmental methods
Forecasts generated subjectively by the forecaster
Educated guesses
 Quantitative methods – based on mathematical modeling
Forecasts generated through mathematical modeling.

13
 Qualitative Forecasting Methods
 The qualitative (or judgmental) approach can be
useful in formulating short-term forecasts and can also
supplement the projections based on the use of any of
the quantitative methods.
 Individual Expert: individual market experts can be
hired to watch for industry trends, perhaps even by
geographic area, and might even work with sales
people to estimate future demand for products
 Executive Opinions/Group Consensus: The subjective
views of executives or experts from sales, production,
finance, purchasing, and administration are averaged
to generate a forecast about future sales.
14
 Usually this method is used in conjunction with some
quantitative method, such as trend extrapolation.
Delphi Method: this method requires one person to
administer and coordinate the process and poll the team
members (respondents) through a series of sequential
questionnaires.

Consumer Surveys: Some companies conduct their own

market surveys regarding specific consumer purchases.
Surveys may consist of telephone contacts, personal
interviews, or questionnaires as a means of obtaining
15
data.
Qualitative Methods
Type Characteristics Strengths Weaknesses
Ex ecutiv A group of Good for One person's
e opinion managers meet & strategic or new- opinion can
come up with a product dominate the
forecast
forecasting forecast

Market Uses surveys & Good It can be

research interviews to determinant of difficult to
identify customer customer develop a good
preferences
preferences questionnaire
Delphi Seeks to develop Ex cellent for Time
method a consensus forecasting long- consuming to
among a group of term product develop
experts
demand,
16
 Quantitative Forecasting Techniques
Quantitative analysis typically involves two
approaches: causal models and time-series methods.
Causal/Regression Methods: Causal models establish a
quantitative link between some observable or known
variable (like advertising expenditures, quality, and
competitors) with the demand for some product.

Time Series Forecasting Methods: Time series

forecasting methods are based on analysis of historical
data (time series: a set of observations measured at
successive times or over successive periods). 17
 Time Series forecasting method
They make the assumption that past patterns in
data can be used to forecast future data points.

Naïve or Projection

Simple Moving Average

Weighted Moving Average

Exponential Smoothening

18
 Time Series Models
Forecaster looks for data patterns as
 Data = historic pattern + random variation
Historic pattern to be forecasted:
 Level (long-term average) – data fluctuates around a
constant mean.
 Trend – data exhibits an increasing or decreasing
pattern
 Seasonality – effects are similar variations occurring
during corresponding periods, e.g., December retail
sales. Seasonal can be quarterly, monthly, weekly, daily,
or even hourly indexes.
19
 Cycle – are the long-term swings about the trend line.

They are often associated with business cycles and may

extend out to several years in length.
Irregular variations - caused by unusual
circumstances

Random variations - caused by chance, cannot be

predicted

20
Time Series Patterns

21
 Time Series Models
1. Naïve or Projection: The easiest time series method
simply projects future demand based on the last period’s
demand. The forecast for the period t, Ft, is simply a
projection of previous period t-1 demand, At-1
Ft=At-1
E.g. If the actual demand of period t is 120, then the
forecast of the period t+1 is 120.
This method, although easy to use, doesn’t make use of
data that is easily available to most managers; thus,
using more of the historical data should improve the
forecast. Averages of past demand might be more
useful and are discussed next.
22
2. Simple Moving Average (SMA): The simple moving
average forecast makes use of more of the historical
demand data than just the last period’s demand.
 An n- period moving average uses the last n periods of
demand as a forecast for next periods demand:

Where :- n = total number of periods in the average

Ft= Forecast for period t
At-1, At-2,….At-n = Actual occurrences for previous
periods ( 1, 2, …,n)

23
The key decision here is how many periods should be
considered in the forecast; higher value of n - greater
smoothing, lower responsiveness, and lower value of n - less
smoothing, more responsiveness.
 The more periods (n) over which the moving average is
calculated, the less susceptible the forecast is to random
variations, but the less responsive it is to changes.
A large value of n is appropriate if the underlying pattern of
demand is stable.
A smaller value of n is appropriate if the underlying pattern
is changing or if it is important to identify short-term
fluctuations.

24
 Simple Moving Average (A company sells storage shed,
Determine the forecast of January using 3 month simple
moving average.)

25
 Forecasting Techniques
3. Weighted Moving Average(WMA): A weighted moving
average is a moving average where each historical demand
may be weighted differently.

One shortcoming of the simple moving average is the

equal weighting of data.

This runs counter to ones intuition that the most recent

data is the most relevant. Thus, the weighted moving
average allows for more emphasis to be placed on the
most recent data. 26
This forecast is:

Where:- wt-1 is the weight applied to the demand

incurred period t-1, and so on. Intuitively, the
expectation would be that the more recent demand
data should be weighted more heavily than older data;
so, generally, one would expect the weights to follow
the relationship wt ≥ wt-1 ≥ wt-2 ≥ ….
The sum of the weights is one.

27
Weighted Moving Average: Consider the weights 3/6,
2/6,1/6 for periods t-1, t-2 and t-3 respectively which
are added to one. Determine the forecast of January.

28
4. Exponential smoothing: Nice properties of a weighted
moving average would be one where the weights not only
decrease as older and older data are used, but one where
the differences between the weights are “smooth”.
Obviously the desire would be for the weight on the most
recent data to be the largest.
The weights should then get progressively smaller the
more periods one considers into the past.
The exponentially decreasing weights of the basic
exponential smoothing forecast fit this bill nicely.

29
 The forecast equation is given by:

30
 Exponential Smoothing Example
Predicted demand = 142 Ford Mustangs
Actual demand = 153
Smoothing constant a = .20

New forecast = 142 + .2(153 – 142)

= 142 + 2.2
= 144.2 ≈ 144 cars
Why use exponential smoothing?
Uses less storage space for data
More accurate
Easy to understand
Little calculation complexity
There are simple accuracy tests

 The smoothing constant α expresses how much our forecast will

react to observed differences.
 If α is low: there is little reaction to differences.
 If α is high: there is a lot of reaction to differences. 32
Selecting Smoothening Constant (α):

The exponential smoothing approach is easy to use and

has been applied successfully by banks, manufacturing
companies. wholesalers, and Other organizations.

The appropriate value of the smoothing constant, α,

however, can make the difference between an accurate
forecast and an inaccurate forecast.

In picking a value for the smoothing constant, the

objective is to obtain the most accurate forecast.
33
Several values of the smoothing constant may be tried, and
the one with the lowest mean absolute deviation (MAD)
could be selected. This is analogous to how weights are
selected for a weighted moving average forecast.
Some forecasting software will automatically select the best
smoothing constant. QM for Windows will display the
MAD that would be obtained with values of α ranging from
0 to 1 in measurements of 0.01.

Exponential smoothing: Main idea: The prediction of the

future depends mostly on the most recent observation, and
on the error for the latest forecast.
To determine a forecast of a given period the forecast of the
previous period should be know .
34
35
36
Selecting the Right Forecasting Model
1. The amount & type of available data
 Some methods require more data than others
2. Degree of accuracy required
 Increasing accuracy means more data
3. Length of forecast horizon
 Different models for 3 month vs. 10 years
4. Presence of data patterns
 Lagging will occur when a forecasting model meant
for a level pattern is applied with a trend

37
 produ c t
t i ng during
 Forecas
life cycle
38
 Measuring Forecast Error
Forecasts are never perfect
Need to know how much we should rely on our chosen
forecasting method.
Measuring forecast error:

E t  A t  Ft
Note that:
Over-forecasts = negative errors.
Under-forecasts = positive errors.
Large values of negative or positive errors shows there
is bias in the forecast.

39
  Mean Absolute Deviation (MAD)
measures the total error in a forecast without regard to
sign
Cumulative Forecast Error (CFE)
Also called running sum of forecast error (RSFE)
Measures any bias in the forecast
Mean Square Error (MSE)
Penalizes larger errors

n
 
n
 t t
2
 A t - Ft A - F
MAD = t =1
MSE = t =1 RMSE = MSE
n n
n
CFE   actual  forecast  RSFE  (At  Ft )
i 1

 Ideal values =0 (i.e., no forecasting error)

Measuring Accuracy: Tracking signal
 The tracking signal (TS) is a measure of how often our
estimations have been above or below the actual value. It is
used to decide when to re-evaluate using a model.
n
RSFE
RSFE  (At  Ft ) TS 
i1 MAD
 Positive tracking signal: most of the time actual values are
above our forecasted values
 Negative tracking signal: most of the time actual values are
below our forecasted values
Usually 3 ≤ TS ≥ 8, out of this range investigate!
Measuring Forecast Accuracy and Error
Weighted (n=3,
Simple t-1=0.45, Exponential Exponential Exponential
S.N Actual Naïve
(n=3) t-2=0.35, (α=0.1) (α=0.5) (α=0.8)
t-3=0.2)
1 110 105 105 105
2 100 110.0 105.5 107.5 109.0
3 120 100.0 105.0 103.8 101.8
4 140 120.0 110.0 108.5 106.5 111.9 116.4
5 170 140.0 120.0 115.0 109.8 125.9 135.3
6 150 170.0 143.3 137.0 115.8 148.0 163.1
7 160 150.0 153.3 152.5 119.2 149.0 152.6
8 190 160.0 160.0 161.0 123.3 154.5 158.5
9 200 190.0 166.7 161.5 130.0 172.2 183.7
10 190 200.0 183.3 178.5 137.0 186.1 196.7
11 190.0 193.3 193.5 142.3 188.1 191.3
MAD 17.8 23.3 26.6 38.4 18.1 16.6
CFE 80.0 163.3 186.0 372.9 166.1 107.9
RMSE 58.3 74.5 81.8 141.5 72.5 61.1
TS 4.50 7.00 7.00 9.71 9.17 6.52
42
 Summary
 Three basic principles of forecasting are: forecasts are rarely
perfect, are more accurate for groups than individual items, and
are more accurate in the shorter term than longer time horizons.
 The forecasting process involves five steps: decide what to
forecast, evaluate and analyze appropriate data, select and test
model, generate forecast, and monitor accuracy.
 Forecasting methods can be classified into two groups:
qualitative and quantitative. Qualitative methods are based on
the subjective opinion of the forecaster and quantitative
methods are based on mathematical modeling.
43
 Time series models are based on the assumption that all

information needed is contained in the time series of data.

 Causal models assume that the variable being forecast is

related to other variables in the environment.

 There are four basic patterns of data: level or horizontal, trend,

seasonality, and cycles. In addition, data usually contain

random variation. Some forecast models used to forecast the
level of a time series are: naïve, simple mean, simple moving
average, weighted moving average, and exponential smoothing.
Separate models are used to forecast trends and seasonality.
44
 Three useful measures of forecast error are mean absolute
deviation (MAD), mean square error (MSE) and tracking
signal.
 There are four factors to consider when selecting a model:
amount and type of data available, degree of accuracy
required, length of forecast horizon, and patterns present in
the data.

45
 re e
T h
t e r
hap
o f C
nd A
E Q &

46