DEMAND FORECASTING

 Forecast is an estimate of future events and trends and is arrived at by systematically

combining past data and projecting it forward in a predetermine a manner.
 For short-term decisions in production planning, distribution etc and selling individual
products would require short-term forecast, upto one-year time horizon, which must be
fairly accurate for specific product items.
 For long-term planning, time horizon being four to five years, information required from
demand analysis would be for broad product groups for facilitating choice of technology,
machine tools, location, etc.
 To survive in the market, management have to be forward-looking and carry out market
and demand analyses of projects and develop strategic business policies.
 Forecasting is a key driver of virtually every design and planning decision made in both an
enterprise and a supply chain

Forecasting- Factors
Companies need to first
 Identify the factors that influence the future demand, and then
 Ascertain the relationship between these factors and future demand

Some of the factors that need to be looked into

 Past demand
 Lead time of products
 Planned advertising or marketing efforts
 State of economy
 Planned price discounts
 Action competitors have taken

Forecasting-Components
 Trend
o Refers to gradual, long term, upward or downward movement in the data over time
 Changes in income, population etc.
 Seasonality
o Refers to short term fairly regular variations related to factors such as weather,
holidays, vacations etc.
o Variations can be daily, weekly or monthly
 Cycles
o Wave like variations of more than one year’s duration or which occur every year
 Business cycle related to economic, political or agricultural conditions
 Random variations
 o Residual variations which are blips in the data caused by chance and unusual

situations

Forecasting Methods:

Qualitative Method
 Qualitative forecasting methods are primarily subjective and rely on human judgment
 Most appropriate when there is little historical data available or when experts have
market intelligence that is critical in making forecast
 Used to forecast future demand for long term in a new industry

Time Series
 Use historical demand to forecast
 Method appropriate when the demand pattern does not vary significantly from one year
to the next

Causal
 Method assumes that the demand forecast is highly correlated with certain factors in
the environment such as, State of economy, interest rates etc.
 Used to determine the impact of price promotions on demand

Simulation
 Methods imitate the consumer choices that give rise to demand to arrive at a forecast
 Simulation is used to combine time series and causal methods to find answers to
 Impact of price promotion, competitors’ stores coming up in the vicinity etc.
 Forecast demand for higher fare seats when there are no seats available at economy
class fare
 Modeling makes use of computers
Forecasts in time series methods based on averages smoothened through averaging
Three techniques used for Averaging
 Naive Forecasts
o Simplest method
o Assumption of demand for the next period based on the actual demand in the
most recent period
 Moving Average method
o Simple moving average
o Weighted moving average
 Exponential Smoothing Models
o Simple moving average
o Weighted moving average

Measures of Forecast Accuracy:

 Mean Absolute Deviation (MAD)
 Mean Square Error (MSE)
 Mean Absolute Percentage Error (MAPE)
 BIAS
 Tracking Signal

SIMPLE MOVING AVERAGE METHOD:

A simple moving average (SMA), is calculated by taking the arithmetic mean of a given set of
values over a specified period. A set of numbers, or prices of stocks, are added together and then
divided by the number of prices in the set. The formula for calculating the simple moving average
of a security is as follows:

SMA =  in=1/n =(D1+D2+D3……Dn)/n,

• where , n=the chosen number of periods,
• i= 1 is the oldest period in the n-period average
• i= n is the most recent period
• D1= the demand in the ’i’ th period

Weighted Moving Average Method:

– A weighted average of past sales is the forecast for the next time period
– A WMA allows for varying, not equal weightage of old demands

– WMA=  in=1 Ci Di
–
 • where Di is the demand during time period ‘i’, Ci is the weight given to that
demand and ‘n’ is the chosen number of periods
• Also 0 Ci  1 , and  in=1 Ci =1

Forecasting Error is simply the difference between the forecast and actual demand for a given
period
• et = Ft – Dt ,
• where et = forecast error for the period t,
• Dt = actual demand for period t, and Ft = the forecast for the period t

Absolute Error (ME) = | et |

Mean Absolute Deviation (MAD) = 1/n ∑nt=1 | et |

• MAD is merely the average error for each forecast.
• Popular because it is easy to understand

Mean Squared Error (MSE) = 1/n ∑nt=1 et2

• Used as an estimate of the variance of the random error et which is σ2

Mean Absolute Percentage Error (MAPE)

=1/n ∑ⁿ t=1 ( |et| / Dt) X100
MAPE is useful for putting forecast performance in the proper perspective

Bias Shows whether the forecast consistently under- or overestimates demand; should fluctuate
around 0
biasn = Sum(t=1 to n)[Et]

Tracking signal
Should be within the range of +6
Otherwise, possibly use a new forecasting method
TSt = Bias / MADt

Case:
The demand for the 10- periods are given as follows:
Determine the forecast for the 11th period using
• 3- Moving Average method
• Exponential Smoothing method (Use Initial forecast= D1 and Smoothing constant = 0.7)

Period (Years) Demand (D)

1 22
 2 26
3 20
4 15
5 21
6 26
7 30
8 35
9 40
10 36
3-MOVING AVERAGE Method

Forecast Absolute
Period Demand 3-MA Square
Error Deviation
(Years) (D) Forecast (F) Error = E2
E= F-D = IEI
1 22 * * *
2 26 * * *
3 20 * * *
4 15 22.67 7.67 7.67 58.78
5 21 20.33 -0.67 0.67 0.44
6 26 18.67 -7.33 7.33 53.78
7 30 20.67 -9.33 9.33 87.11
8 35 25.67 -9.33 9.33 87.11
9 40 30.33 -9.67 9.67 93.44
10 36 35.00 -1.00 1.00 1.00
45.00 381.67

Forecast for 11th Period= F11= (35+40+36)/3= 37

MAD= 45.00 / 7 = 6.43

MSE = 381.67 / 7= 54.52

EXPONENTIAL SOOMETHING Method

Exponential smoothing is generally used to make short term forecasts, but longer-term forecasts
using this technique can be quite unreliable. More recent observations given larger weights by
exponential smoothing methods, and the weights decrease exponentially as the observations
become more distant. When the parameters describing the time series are changing slowly over
time, then these methods are most effective.
Exponential smoothing of time series data assigns exponentially decreasing weights for newest
to oldest observations. In other words, the older the data, the less priority (“weight”) the data is
given; newer data is seen as more relevant and is assigned more weight. Smoothing parameters
(smoothing constants)— usually denoted by α— determine the weights for observations.

Given:
Initial Forecast = D1= 22
Smoothing Constant =α = 0.7

Formula:
Ft= Ft-1 + (Dt-1 - Ft-1)
F2= F1 + (D1-F1)

Forecast Absolute Square

Period Demand 3-MA
Error Deviation Error =
(Years) (D) Forecast (F)
E= F-D = │E│ E2
1 22 22.00 0.00 0.00 0.00
2 26 22.00 -4.00 4.00 16.00
3 20 24.80 4.80 4.80 23.04
4 15 21.44 6.44 6.44 41.47
5 21 16.93 -4.07 4.07 16.55
6 26 19.78 -6.22 6.22 38.69
7 30 24.13 -5.87 5.87 34.41
8 35 28.24 -6.76 6.76 45.70
9 40 32.97 -7.03 7.03 49.39
10 36 37.89 1.89 1.89 3.58
F11 = 36.57 47.07 268.83

Forecast for 11th Period= F11= F10+ α (D10-F10) = 37.89 + 0.7 (36- 37.89) = 36.57
MAD= 47.07/ 10 = 4.707
MSE = 268.83 / 10= 26.883

As MAD & MSE values are less in exponential smoothing method, it should be preferred
over 3-MA forecasting method.