Aditya Pal

IIT DELHI

https://t.me/adi20pathshala
FORCASTING
 DEMAND forecast is basically concerned with the estimation of
DEMAND.
 Time Horizons in forecasting

Intermediate Long term

Short term
forecast forecast
forecast

• Time period btw 1 • They are calculated • Time period 2,3……years.

to 3 month. for aggregate • new type of product, major
• They are calculated product. capital investment.
for specific product. • Application: • QUALITATIVE techniques of
a. Purchasing, • sales planning & forecasting are used
overtime decisions sales force decisions, • Application: Capital
inventory planning, planning, plant location,
b. machine new product planning, etc,.
maintenance, etc., etc,.
IES 2007
Which one of the following is not a purpose of long-term
forecasting?
(a) To plan for the new unit of production
(b) To plan the long-term financial requirement.
(c) To make the proper arrangement for training the personnel.
(d) To decide the purchase programme.
 Methods and types of forecasting

Quantitative
Qualitative

Survey Group Group Delphi

averaging Consensus
Group averaging
• This involves averaging of individual forecast to obtain one value.
• This method is fast simple and inexpensive
Group consensus
• This method involves all the subject expert meeting together at one
place for obtaining one unanimous forecast.
Delphi
• This is one of the most widely used technique in the
industry. The coordinating industry asked the subject
expert to give their individual forecast along with
supporting information. Which is used to arrive at
those forecast value.
• , pooling of expert opinions
IES-1995
Which one of the following methods can be used for forecasting the
sales potential of a new product?
(a) Time series analysis
(b) Jury of executive opinion method
(c) Sales force composite method
(d) Direct survey method
IAS-1996
For sales forecasting, pooling of expert opinions is made use of in
(a)Statistical correlation
(b) Delphi technique
(c) Moving average method
(d) Exponential smoothing
Quantitative forecasting
• This method is generally used for short and medium
term forecasting.

 Time series model :-In this method the variable that is

being forecast is decomposed into its components (N)

 Associative model:-
IES-2001
Assertion (A): Time series analysis technique of sales-forecasting
can be applied to only medium and short-range forecasting.
Reason (R): Qualitative information about the market is necessary
for long-range forecasting.
(a) Both A and R are individually true and R is the correct explanation of
A
(b) Both A and R are individually true but R is not the correct
explanation of A
(c) A is true but R is false
(d) A is false but R is true
 Trend Cyclical

Seasonal Random
Trend Component Seasonal Component

0 5 10 15 20
M T W T F
Cyclical Component
Random Component
IES-1997
Given T = Underlying trend, C = Cyclic variations within the trend, S =
Seasonal variation within the trend and R = Residual, remaining or
random variation, as per the time series analysis of sales
forecasting, the demand will be a function of:
(a) T and C (b) R and S
(c) T, C and S (d) T, C, S and R
Rolling horizon in forecast is used for easy updating of
changes and maintaining same length of forecast horizon
by adding a new period when one period is over
IES 2012
Rolling horizon in forecast is used for
(A) Allowing same length of forecast horizon by easily adding a new
period when one period is over
(B) Easy updating of changes and maintaining same length of forecast
horizon by adding a new period when one period is over
(C) Easy updating of changes and there is no addition of a new period
(D) Different reasons other than the above
Time series analysis technique of sales-forecasting can be applied to
only medium and short-range forecasting.

Qualitative information about the market is necessary for long-range

forecasting.
Simple moving average
• Simple moving average: -Moving Average obtained by adding and
averaging the value from a given number of period repeatedly, each
time deleting the oldest value and adding a new value.
Simple moving average
If N is the period of moving average

S.No Demand Ft (for N=3)

1 100
2 120
3 110
4 130
5 140
6 160
7 150
8 140
9
IES 2013
If N is the period of moving average the number of demand data to
be stored for calculating the moving average for a period is the
demand of last________periods.
(A)(N+1) periods
(B) (N-1) periods
(C) (N) periods
(D) (N-2) periods
IES 2011
Consider the following characteristics.
(1) It is very sensitive to small movements in the data
(2) The technique is simple
(3) The method is affected by the personal prejudice of the people
Which of these characteristics of moving average method of forecasting
are correct?
(A) 1, 2 and 3 (B) 1 and 2 only (C) 2 and 3 only (D) 1 and 3 only
Weighted moving average
• In this method the weights are given to the values
vary,

 The highest weight is given to the latest value and the

weight decreases as the values become old.
 Weighted moving average
S.No Demand Ft (for N=4)
1 100
2 120
3 110
4 130
5 140
6 160
7 150
8 140
9
SOD method to find the weight SUM OF NATURAL
N/SMN, N-1/SMN, N-2/SMN, N-3/SMN, ----------- NUMBER=SMN
Exponential smoothing

Ft+1 =  Dt + (1- ) Dt - 1 + (1- )2Dt - 2 + ...

Exponential Smoothing
• Assumes the most recent observations have the highest
predictive value
• The weightage of the data diminishes exponentially as the
data become older
• gives more weight to recent time periods

Ft+1 = Ft + (Dt - Ft)

et
Ft+1 = Forecast value for time t+1
Dt = Actual value at time t
 = Smoothing constant
 α

α(1- α)

Weightage α(1- α)2

Of
Past data α(1- α)3

α(1- α)t-3
α(1- α)t-2

α(1- α)t-1

1 2 3 t-3 t-2 t-1 t

period
Forecast Effects of
Smoothing Constant 
Ft+1 = F +  (Dt - Ft)
t

or Ft+1 =  Dt + (1- ) Dt - 1 + (1- )2Dt - 2 + ...

w1 w2 w3
Weights

=
Prior Period 2 periods ago 3 periods ago
 (1 - ) (1 - )2

= 0.10 10% 9% 8.1%

= 0.90 90% 9% 0.9%

IES 2012
In an exponentially weighted moving average, the weight of the
demand of past periods
(A)Increases as age of the data increases
(B) Increases as age of the data deceases
(C) Decreases as age of the data increases
(D) Has no relationship with age of the data
If α is not given then α=2/N+1
Values of α varies 0 to 1

F1 value will be given.

If value is not given, take F1=D1
Another case average of all
demand=F1.
Exponential smoothening methods are best suited
under conditions when forecasting horizon is
relatively large
Exponential Smoothing – Example 1
Ft+1 = Ft + (Dt - Ft)
i Di

Week Demand
1 820 Given the weekly demand
2 775 data what are the exponential
3 680 smoothing forecasts for
4 655 periods 2-10 using =0.10?
5 750
Assume F1=D1
6 802
7 798
8 689
9 775
10
Exponential Smoothing – Example 1
Ft+1 = Ft + (Dt - Ft)
i Di

Week Demand
1 820
2 775
3 680
4 655
5 750
6 802
7 798
8 689
9 775
10
IES 2014
Exponential smoothening methods are best suited under conditions
when
(A) forecasting horizon is relatively large
(B) forecasting for large number of items
(C) available outside information is more
(D) All of the above
IES-1999
A company intends to use exponential smoothing technique for
making a forecast for one of its products. The previous year's
forecast has been 78 units and the actual demand for the
corresponding period turned out to be 73 units. If the value of the
smoothening constant α is 0.2, the forecast for the next period will
be:
(a) 73 units (b) 75 units (c) 77 units (d) 78 units
IES-2005
For a product, the forecast for the month of January was 500 units.
The actual demand turned out to be 450 units. What is the forecast
for the month of February using exponential smoothing method
with a smoothing coefficient = 0.1?
(a) 455 (b) 495 (c) 500 (d) 545
Responsiveness & stability

time
• Responsiveness indicates that forecast as calculated
have a fluctuating or swinging pattern.
 Stability means that the forecast show a leveled or flat
character as the value of N increases the forecast become
stable.

 Lower value of N results in forecast being more

responsive
Ft+1 = Ft + (Dt - Ft)
If =0 Ft+1 = Ft ---------------- STABLE

If =1 Ft+1 = Dt ---------------- RESPONSIVE

• Higher the value of α, more responsive the forecast
will be and this is desirable for forecasting of new
products.

 Whereas lower value of α makes the forecast more

stable and this desirable for old and stable products.
IES-2008

Using the exponential smoothing method of forecasting, what will

be the forecast for the fourth week if the actual and forecasted
demand for the third week is 480 and 500 respectively and α = 0·2?
(a) 400 (b) 496 (c) 500 (d) 504

F4 = α d3 + 1 −α F3 = 0.2 480) + (0.8)500 = 96 + 400 = 496

Error analysis
• It is assumed that the forecasting model should over
estimate and under estimate with equal magnitude so
that the errors produces by forecasting model will fit
into a Normal distribution curve
Measuring Forecast Error

• Forecasts are never perfect

• Need to know how much we should rely on
our chosen forecasting method
• Measuring forecast error:
Et = Dt − Ft

• Note that over-forecasts = negative errors

and under-forecasts = positive errors
 Measuring Forecasting Accuracy

• Mean Absolute Deviation (MAD)

• measures the total error in a forecast
without regard to sign

• Cumulative Forecast Error (CFE)

• Measures any bias in the forecast

• Mean Square Error (MSE)

• Penalizes larger errors

• Tracking Signal
• Measures if your model is working
• Good tracking signal has low values
Measuring Forecasting Accuracy
Mean Error or BIAS
IES-2009
Which of the following is the measure of forecast error?
(a) Mean absolute deviation
(b) Trend value
(c) Moving average
(d) Price fluctuation
IES-2004
It is given that the actual demand is 59 units, a previous forecast 64
units and smoothening factor 0.3. What will be the forecast for next
period, using exponential smoothing?
(a) 36.9 units (b) 57.5 units (c) 60.5 units (d) 62.5 units
IES 2007
Consider the following statements:
Exponential smoothing
1. Is a modification of moving average method
2. Is a weighted average of past observations
3. Assigns the highest weight age to the most recent observation
Which of the statements given above are correct?
(a) 1, 2 and 3 (b) 1 and 2 only
(c) 2 and 3 only (d) 1 and 3 only
Associative Forecasting

Used when changes in one or more independent

variables can be used to predict the changes in the
dependent variable

Most common technique is linear

regression analysis

We apply this technique just as we did in the

time series example
Linear regression in forecasting

Linear regression is based on

1. Fitting a straight line to data
2. Explaining the change in one variable through changes in
other variables.

dependent variable = a + b  (independent variable)

Example: do people drink more when it’s cold?

Alcohol Sales

Which line best

fits the data?

Average Monthly
Temperature
IES-2008
Which one of the following is not a technique of Long Range
Forecasting?
(a)Market Research and Market Survey
(b)Delphi
(c) Collective Opinion
(d) Correlation and Regression
NOTE:-
• Regression will forecast a higher value compared to moving average
method.
IES-2005
Which one of the following forecasting techniques is most suitable
for making long range forecasts?
(a) Time series analysis (b) Regression analysis
(c) Exponential smoothing (d) Market Surveys
IES-2005
Which one of the following methods can be used for forecasting
when a demand pattern is consistently increasing or decreasing?
(a) Regression analysis (b) Moving average
(c) Variance analysis (d) Weighted moving average
IES-2003
Which one of the following statements is correct?
(a) Time series analysis technique of forecasting is used for very long
range forecasting
(b) Qualitative techniques are used for long range forecasting and
quantitative techniques for short and medium range forecasting
(c) Coefficient of correlation is calculated in case of time series
technique
(d) Market survey and Delphi techniques are used for short range
forecasting
IES-2009
• Assertion (A): Moving average method of forecasting demand gives
an account of the trends in fluctuations and suppresses day-to-day
insignificant fluctuations.
• Reason (R): Working out moving averages of the demand data
smoothens the random day-to-day fluctuations and represents only
significant variations.
(a) Both A and R are true and R is the correct explanation of A
(b) Both A and R are true but R is NOT the correct explanation of A
(c) A is true but R is false
(d) A is false but R is true
IES-2006
Which one of the following is a qualitative technique of demand
forecasting?
(a) Correlation and regression analysis
(b) Moving average method
(c) Delphi technique
(d) Exponential smoothing
IES-2006
Which one of the following statements is not correct for the
exponential smoothing method of demand forecasting?
(a) Demand for the most recent data is given more weightage
(b) This method requires only the current demand and forecast
demand
(c) This method assigns weight to all the previous data
(d) This method gives equal weightage to all the periods
 TYPES OF FORECASTS
PASSIVE FORECASTS
Where the factors being forecasted
are assumed to be constant over a
period of time and changes are
ignored.
ACTIVE FORECASTS
Where factors being forecasted are
taken as flexible and are subject
to changes.
IES 2013
Forecasting which assumes a static environment in the future is:
(A)Passive forecasting
(B) Active forecasting
(C) Long term forecasting
(D) Short term forecasting