DEMAND FORECASTING

UNIT - III
SO
WHAT

“IS”

DEMAND

FORECASTING?
 Demand Forecasting
Demand forecasting is a technique of
predicting or estimating demand in future on
the basis of the behaviour of factors which
affect the demand.
It is just not simple guessing game but it
involves use of various scientific techniques
plus proper judgement & acumen on the parts
of decision-makers.
• Demand forecasting is a specific type of
forecasting, which enables the manager to
minimize elements of risk and uncertainty
• The likely future event has to be given form and
content in terms of projected courses of variable,
i.e. is forecasting.
• The manager can conceptualize the future in
definite terms.
• Forecasting customer demand for
products and services is a proactive
process of determining what products are
needed where, when, and in what
quantities. Consequently, demand
forecasting is a customer–focused activity.
Factors involved in Demand Forecasting
1. How far ahead
i) Short run
ii) Long run

2. Undertaken at three levels:

a. Macro-level
b. Industry level eg., trade associations
c. Firm level
3. Should the forecast be general or specific (product-wise)?
4. Problems or methods of forecasting for “new” vis-à-vis “well
established” products.
5. Classification of products – producer goods, consumer durables,
consumer goods, services.
6. Special factors peculiar to the product and the market – risk and
uncertainty. (eg., ladies’ dresses, political stability)
 Features
• It is in terms of specific quantities
• It is undertaken in an uncertain atmosphere.
• A forecast is made for a specific period of time which
would be sufficient to take a decision and put it into action.
• It is based on historical information and the past data.
• It tells us only the approximate demand for a product in
the future.
• It is based on certain assumptions.
• It cannot be 100% precise as it deals with future expected
demand
Significance of Demand Forecasting
• Production Planning.
• Sales forecasting.
• Control of business.
• Inventory control.
• Pricing Policy
• Stability.
• Labor requirement
• Growth and long-term investment programmes.
• Economic planning and policy making
 SCOPE OF DEMAND
FORECASTING
Levels of forecasting
-- Macro level
-- Industry level
-- Firm level
In macro level, it takes into account the aggregates such as
NI, expenditure, IIP etc., while estimating demand.

At industry level, the forecasting is made for the whole

industry.

At the firm level it involves forecasting the firm’s share in

the total market demand
DETERMINANTS FOR DEMAND
FORECASTING

1. Capital goods – Goods required for further

production of goods
Demand for capital goods is derived
demand
- Replacement demand
- New demand
2. Durable consumer goods— Goods used
continuously for a period of time
1. Buy vs. Replacement decision
2. Family Characteristics
3. Special attached facilities
4. Prices & credit facilities
3. Non-durable consumer goods—
Commodities which are used in a single
act of consumption
Demand for these goods is
influenced by
- Disposable income of
people/Purchasing Power
- Price of the commodity
- Size and characteristics of population/
Demography
 LENGTH OF FORECASTING
1. Short Term: up to 12 months, determine sales
quota, inventory control, production schedules,
budgeting & planning cash flows.
2. Medium Term: from 1 – 2 years, determine
the rate of maintenance, schedule of operation &
budgetary control over expenses.
3. Long Term: from 3 – 10 years, determine
apital expenditures, personnel requirement,
financial requirements, raw material
requirements & scope of R&D programmes.
Short-term Forecasting Purposes:
• Production scheduling
• Evolving a sales policy.
• Determining price policy.
• Evolving a purchase policy of raw material.
• Fixation of sales targets & incentives.
• Determining Short-term Financial Planning.
• Evolving suitable advertising & promotion
program
 Long-Term Forecasting
• Business Planning for new unit or
expansion of an existing unit.
• Man power Planning.
• Long-term financial planning.
CRITERIA FOR GOOD DEMAND
FORECASTING
1. Accuracy
2. Plausibility
3. Durability
4. Availability
5. Economy
6. Simplicity & ease of comprehension
 FORECASTING DEMAND FOR
NEW PRODUCTS
Joel Dean suggested the followings for
forecasting demand for new products.

• Project the demand for the new product as

an outgrowth of an existing old product
• Analyse the new product as a substitute for
some existing old product
• Estimate the rate of growth & ultimate level of
demand for the new product on the basis of the pattern
of growth of established products
• Estimate the demand by making direct enquiries from
the ultimate purchasers, either by the use of samples
or on a full scale.
• Offer the new product for sale in a sample market.
• Survey consumer reaction to a new product indirectly
through the eyes of specialised dealers who are
supposed to be informed about consumers’ need &
alternative opportunities.
 PRESENTATION OF A
FORECAST TO THE MGMT
1. Make the forecast as easy for the mgmt to
understand as possible
2. Avoid using vague generalities
3. Always pin point major assumptions & sources
4. Give the possible margin error
5. Avoid making undue qualification
6. Omit details about methodology and calculation
7. Make use of charts and graphs as much as
possible for easy comprehension.
METHODS OF DEMAND
FORECASTING
 CONSUMER SURVEY METHOD
• Least sophisticated method
• Customers are directly contacted to find out their
intentions to buy commodities in the near future –
usually for one year.
• Most useful when bulk of the sales is made to the
industrial producers.
• Intentions recorded through personal interviews, mail or
post service, telephone interviews and questionnaires.
• Three types of Consumer Survey
– Complete Enumeration Method
– Sample Survey Method
– End use Method or Input-Output
 Sales Force Opinion Method
• The salesmen are questioned & their response or
reactions aggregated.
• This method is very cheap & easy.
• It has the advantage of first hand knowledge of the
salesmen.
• This method is quite useful for forecasting
demand for new products.
• This method is otherwise known as “Reaction
survey” method.
Disadvantage of Sales force Opinion
• Salesmen could generally understand the
situations only near-future forecasting, therefore it
is useful for a short period.
• Salesmen are ignorant of broader economic
changes in the market which has to consider while
forecasting.
• Salesmen can be affected by either by congenital
optimism or congenital pessimism.
 EXPERT OPINION
• Here experts in the field has been asked for
estimating their likely sales.
• Experts include executives directly involved
in the market, such as distributor, dealers,
suppliers, industry analyst, specialist mktg
consultants, trade associations officers.
• Each expert is asked independently to
provide confidential estimate & results
could be averaged.
 DELPHI METHOD
• It is developed at Rand Corporation of the
USA in the late 1940s.
• It was developed by Olaf Helmer, Dalkey, &
Gordon
• The forecasters are given the forecasts and
assumptions of other experts, and a final report
is compiled with the combined consensus of
the experts.
• It is more popular in forecasting non-economic
rather than economic variables
Advantages:

 Facilitates the maintenance of anonymity of

the respondent’s identity through out the
course.

 This technique saves time & other resources

 MARKET SURVEY METHOD
• CONTROLLED EXPERIMENTS
Different determinants of demand are varied and price quantity
relationships are established at different points of time in the same
market or different markets.
Only one determinant varied ; others kept constant.
• SIMULATED MARKET SITUATION
An artificial market situation is created and “consumer clinics”
selected. Consumers are asked to spend time in an artificial
departmental store and different prices are set for different buyer
groups.
The responses to the price changes are observed and necessary
decisions taken.
 Limitation of Controlled
Experiments
• Expensive & time consuming
• Risky because they may lead to unfavorable
reaction on dealers, consumers, &
competitors
• Difficulty in planning the study
• Difficult to satisfy the condition of
homogeneity of market.
 Fitting Trend Line by Observation
• It is easy and quick method to project the demand.
• It involves the plotting of annual sales on a graph
and then estimating just by observation where the
trend line lies. The line can be simply extended to
a future period and corresponding sales forecast
read against that year.
• As this methods lacks scientific temper, it could
not able to estimate in detail.
 Least Square Method
• Based on analysis of past sales patterns
• Shows effective demand for the product for a
specified time period
• The trend is estimated by using the Least
Square Method.
• This system of forecasting is considered
“naive” because it doesn’t explain the reason
for the change.
A producer of soaps decides to forecast
the next years sales of his product.
The data for the last five years is as
follows:
YEARS SALES IN Rs.LAKHS
1996 45
1997 52
1998 48
1999 55
2000 60
The data is plotted on a graph:
The equation for the straight line trend is
Sales = a + b T (or year no.)
Where a & b are constants representing intercept
& slope respectively.
To determine value of a & b, following two
equations need to be solved.
∑S = Na + b ∑ T (Eq.1)
∑ST = a ∑T + b ∑T2 (Eq.2)

Where N is no of years, months etc for which data is available

 YEARS SALES Rs. T T2 ST
LAKHS (S)
1996 45 1 1 45
1997 52 2 4 104
1998 48 3 9 144
1999 55 4 16 220
2000 60 5 25 300
N=5 ∑S=260 ∑T=15 ∑T2=55 ∑ST=813

Substituting the above values in the normal equations:

260=5a +15b (Eq.3)
813=15a + 55b (Eq.4)
solving the two equations,
a = 42.1 , b = 3.3
Therefore, the equation for the straight line
trend is
S=42.1 + 3.3T
Using this equation we can find the trend values for
the previous years and estimate the sales for the
year 2001 as follows:
Y 1996 = 42.1+3.3(1) = 45.4
Y 1997 = 42.1+3.3(2) = 48.7
Y 1998 = 42.1+3.3(3) = 52.0
Y 1999 = 42.1+3.3(4) = 55.3
Y 2000 = 42.1+3.3(5) = 58.6
Y 2001 = 42.1+3.3(6) = 61.9

Thus, the forecast sales for year 2001 is Rs.61.9 lakhs.

 MOVING AVERAGES METHOD
• A moving average is an average that is updated or recomputed
for every new time period being considered.
• Each MA is based on values covering a fixed time interval,
called ‘period of moving average’ & is shown against the
centre of the period.
• When period of MA is odd, the successive values of moving
averages are placed against the middle value of concerned
group of times. For example, if n=7 the first moving average
value is placed against middle period i.e. 4th value; the second
MA value is placed against time period 5 & so on.
• When period of MA is even, then there are two middle
periods. By using centering technique, a two period avg of the
moving avg is taken & place them in between the
corresponding time period.
 YEAR SALES IN
Rs.LAKHS
1993 12
These are the annual sales
1994 15
of goods during the period
of 1993-2003. 1995 14
We have to find out the 1996 16
trend of the sales using (1) 1997 18
3 yearly moving averages 1998 17
and (2) 4 yearly moving 1999 19
averages and forecast the 2000 20
value for 2005. 2001 22
2002 25
2003 24
3 yearly period:
The value of 1993 + 1994 +1995
12 +15+14 = 41 written at the capital period 1994 of the years
1993, 1994 and 1995
YEAR SALES (Rs. 3 YEARLY 3 YEARLY
LAKHS) MOVING MOVING
TOTAL AVG. TREND
VALUES
1993 12 - -
’94 15 41 41/3= 13.7
’95 14 45 45/3= 15
’96 16 48 48/3 =16
’97 18 51 51/3 =17
’98 17 54 54/3 = 18
’99 19 56 56/3 = 18.7
2000 20 61 61/3 = 20.2
’01 22 67 67/3 = 22.3
’02 25 71 71/3 = 23.7
4 YEARLY MOVING AVERAGES
YEAR. SALES (Rs. 4 YEARLY MOVING TOTAL 4 YEARLY
LAKHS) MOVING TOTAL OF PAIRS OF MOVING AVG.
YEARLY TOTAL TREND VALUES

’93 12 - - -
’94 15 - - -
57
’95 14 120 120/8 = 15
63
’96 16 128 128/8 = 16
65
’97 18 135 135/8 = 16.9
70
’98 17 144 144/8 = 18
74
’99 19 152 152/8 = 19
78
’00 20 164 164/8 = 20.5
86
’01 22 177 177/8 = 22.1
91
’02 25 - -
’03 24 - - -
57 = ‘93 + ‘94 +’95 + ‘96 = 12 + 15 + 14 + 16

120= 57 +63, 128 = 16 +65 and so on.

120 is total of 8 years and so the avg. is calculated by dividing 120 by 8
The trend values from the previous tables can be
plotted on a graph as follows:
Months SALES IN
Rs.LAKHS 4 months moving average
Jan 1056
Feb 1345
Mar 1381 4 months moving average
April 1191 = (1056+1345+1381+1191)/4
May 1259
= 1243.25
Jun 1361
Which will be the forecast for
Jul 1110
the month of May
Aug 1334
Sept 1416
Oct 1282
Nov 1341
Dec 1382
Months SALES IN Average Error
Rs.LAKHS
Jan 1056
Feb 1345
Mar 1381
April 1191
May 1259 1243.25 15.75
Jun 1361 1294 67
Jul 1110 1298 -188
Aug 1334 1230.25 103.75
Sept 1416 1266 150
Oct 1282 1305.25 -23.25
Nov 1341 1285.5 55.5
Dec 1382 1343.25 38.75
 Exponential Smoothing
• It is used to weight data from previous time
periods with exponentially decreasing importance
in the forecast.
• It is one of the popular approach for short term
forecasting.
• Weight assigned to each value reflects degree of
importance of that value.
• More recent values being more relevant for
forecasting, these are assigned greater weight than
previous period values.
• Weights (w) lies between zero & one.
 F t+1 = w. Xt + (1-w) . Ft

Where

F t+1 – the forecast for next time period t+1

F t – the forecast for current time period t
X t – the actual value of present time period
w – a value between 0 < w<1,
‘w’ is referred as exponential smoothing
constant.
This value of ‘w’ is determined by the
forecaster.
 REGRESSION ANALYSIS
• Relationship is established between quantity
demanded being dependent variable and one or
more independent variable such as income, price
of the related goods, price of the commodity under
question, advertisement cost, etc.
• Based on this relationship, the demand trends are
forecasted.
• It can also used when we have more than one
independent variables.
• Principal advantage of this method is that
besides demand forecast, it explains why
demand has been at the level it is.
• It is neither mechanistic like trend method
nor as subjective as the expert opinion
survey method.
• Usually, time series data is used, but we
may use cross section data also.
• As this method is also based on past data,
the forecast will be unrealiable.
 REGRESSION METHOD
“Method of Least Squares”
YEAR 2005 2006 2007 2008 2009
SALES 240 280 240 300 340
(Rs. In
crores)

From the above data we can project the sales for ‘2010, ‘2011, ‘2012.
First we calculate the required values which are (i) Time Deviation,
(ii) Deviation Squares, (iii) Product of time deviation and sales.
YEAR (n) SALES (RS. TIME TD SQUARED (x2) PRODUCT OF
CRORE) (y) DEVIATION TIME
FROM MIDDLE DEVIATION &
YEAR 2007 (x) SALES (xy)

2005 240 -2 4 -480

2006 280 -1 1 -280
2007 240 0 0 0
2008 300 +1 1 +300
2009 340 +2 4 +680
n=5 ∑y = 1400 ∑x = 0 ∑x2 = 10 ∑xy = 220
The equation is
Y = a + bx
‘a’ – independent variable
‘b’ – exhibits rate of growth
a & b can be found out as follows:
a = ∑y / n = 1400 / 5 = 280
b = ∑xy / ∑ x2 = 220/10 = 22
Now, applying values to the regression equation,
Y = 280 + 22x
Hence, sales projection from 2010-2012 can be ascertained.
2010 = 280 + 22 (3) = Rs.346 crores
2011 = 280 + 22 (4) = Rs.368 crores
2012 = 280 + 22 (5) = Rs.390 crores
“Method of Simple linear Regression”
The linear trend can be fitted in the equation

Sales = a + b (Price)

i.e. S = a + bP

where in, a and b are constants.

b = n∑Si Pi- (∑Si)(∑Pi)

n∑Pi2 – (∑Pi) 2

a = ∑Si - b ∑ Pi
n
 e.g. fit a linear regression line to the following data &
estimate the demand at price Rs.30

YEAR ’81 ’82 ’83 ’84 ’85 ’86 ’87 ’88 ’89 ’90 ’91 ‘92
PRICE 15 15 12 26 18 12 8 38 26 19 29 22
(Pi)
SALES 52 46 38 37 37 37 34 25 22 22 20 14
(Si) in
1000 units
 To find the values of a and b the following table is
Pi Si Piconstituted:
2
Si2 Si Pi
15 52 225 2704 780
15 46 225 2116 690
12 38 144 1444 456
26 37 676 1369 962
18 37 324 1369 666
12 37 144 1369 444
8 34 64 1156 272
38 25 1444 625 950
26 22 676 484 572
19 22 361 484 418
29 20 841 400 580
22 14 484 196 308
∑Pi = 240 ∑Si = 384 ∑Pi2 = 5708 ∑Si2 = ∑Si Pi =
13716 7098
b = n∑Si Pi- (∑Si)(∑Pi) = 12(7098)-(240)(384) = 0.641
n∑Pi2 – (∑Pi) 2 2
12 (5708)-(240)

a = ∑Si - b ∑ Pi = [384-(240)(-0.641)] = 44.82

n 12
Thus the regression line is S= 44.82 - 0.641P
By assigning value 30 to P,
The corresponding sales level is
S = 44.82 – 0.641 (30) = 25.29 thousand units
 BAROMETRIC METHOD
• Improvement over trend projection method
• Events of the present are used to predict future
demand
• Basic approach- constructing an index of relevant
economic indicators
 Leading indicators
 Coincident indicators
 Diffusion indices
 Simultaneous Equation Method
• It involves the development of a complete
model which can explain the behaviour of all
the variables which the firm can control.
• The number of equations equals the number of
dependent variables.
• After the model is developed, it is estimated
through some appropriate method such as the
Least Square Method.
• The model is then solved for each of variables.
• It is very costly & time consuming.
 ARIMA Method
(Auto Regressive Integrated Moving Average)
• Otherwise known as Box-Jenkin Technique.
• This method combines smoothing method
with auto regressive method.
• Used for short term forecasting.
• There are five stages of analysis in the
method
 Five Stages of Analysis
• Removal of the Trend
• Model Identification
• Parameter Estimation
• Verification
• Forecasting
1. Removal of the Trend: Useful in those
time series that don’t have long term trend
component. To remove underlying trend
in the time series, method of differencing
is applied. The analysis is performed on
the first differences of successive
observation.
2. Model Identification:
1. The order of involvement of autoregressive
terms
2. Number of differences of the original series
necessary to remove any inherent trend
3. The order of moving average terms
3. Parameter Estimation: Once a particular
combinations of the three elements is
identified, the method of least square is
used to fit this model to the time series.
4. Verification: the goodness of the fit of the
estimated model is checked by analysing
the residuals it generates. If the residuals
don’t show any specific pattern is good fit.
If it is not good fit, we need to repeat the
process by starting afresh from stage 2 &
try to develop a new group of
combinations
5. Forecasting: