2. OBJECTIVES FOR DEMAND
FORECASTING
• Understand the role of demand forecasting
• Identify reasons for demand forecasting
• Study of Forecasting methodologies
• Selecting the right forecasting method.
• Measurement of forecasting errors.
3. INTRODUCTION
Predicting future demand of products/services
of an organisation
Forecast = To estimate/calculate in advance.
Guiding factor- for deciding the capacity and
location of new facility.
The staffing decisions should be in line with
the demand forecasts.
It affects administrative plans and policies.
4. To minimize
Maximize
losses of
gains for To offset
uncontrollabl
actions of the actions
e events
organisation of
competitor
Maximize
gains for
external
Material
environmen REASONS FOR requiremen
t
DEMAND t planning
FORECASTING
To
develop In decision
policies making for
To provide budgeting
To develop adequate
administrativ staff to
e plans support
requirement
s
6. Qualitative Analysis
1) Consumers Survey: Complete Enumeration Method
The forecaster undertakes a complete survey of all
consumers whose demand he intends to forecast.
Once this information is collected, the sales forecasts
are obtained by simply adding the probable demands
of all consumers.
The principle merit of this method is that the
forecaster does not introduce any bias or value
judgment of his own.
But it is a very tedious and cumbersome process; it is
not feasible where a large number of consumers are
involved
7. 2) Consumer Survey-Sample Survey Method
Under this method, the forecaster selects a few
consuming units out of the relevant population and then
collects data on their probable demands for the product
during the forecast period.
The total demand of sample units is finally blown up to
generate the total demand forecast.
Compared to the former survey, this method is less
tedious and less costly, and subject to less data error;
but the choice of sample is very critical.
If the sample is properly chosen, then it will yield
dependable results; otherwise there may be sampling
error.
8. 3) Sales Force Composite
The sales force composite method of forecasting
starts with the forecaster asking for opinions about
future sales from every member of the sales staff
currently working in the field.
Each sales force member states how many sales
she thinks she'll make during the given forecasting
period.
Department managers look over and adjust
salespeople's predictions before turning the
numbers over for forecasting.
Predictions are usually checked against historical
9. 4) Executive Opinion Poll
Forecasters using the executive opinion or expert
opinion method poll executives or experts from
within the company and ask their opinion on the
optional sales for the given forecasting time period.
The forecaster will then average the individual
judgments or try for a group consensus.
Executive opinion polls are often used to verify (or
invalidate) other qualitative methods, especially
sales force composites.
10. 5) Delphi Method
Dis-advantages: Biased , non-response situation , time consuming.
Advantages: No pressure.
11. 6) Past Analogies
Sometimes data on the exact time of a particular
event (or events) are available.
Experts use cases where similar events have
occurred at different times or in different geographic
areas and compare them to the issue at hand.
If occurrence or no-occurrence of an event is on a
regular basis, then the data can be thought of as
having a repeated measurement structure.
It helps to select a large number of similar situations,
rather than basing a decision on comparison with only
one case.
12. Quantitative analysis
Forecast future demand by using quantitative data from
the past and extrapolating it to make forecasts of future
levels.
Demand for existing products can be forecasted by
using this method.
They are used when historical data is available.
There are of two types of techniques
1. Time series analysis
2. Causal analysis
13. Time series analysis
Time series of historical demand data with respect to time intervals
(periods) in the past is used to make predictions for the future demand.
Following are the five popular methods
Simple moving average
Simple exponential smoothing
Holt’s double- exponential smoothing
Winters’ triple- exponential smoothing
Forecasting by Linear regression analysis
14. Simple moving average
It is suitable under situations where there is neither a
growth nor a decline trend shown by the actual past
data for forecasting.
For eg : If we have past data of the actual sales of a
product for the months of Jan, Feb and March, we
take the simple average of these sales figures for
the three months. This simple average becomes the
forecast for the next month i.e April.
15. Simple Moving Average Method
Example : four week moving average
Example: Three Period Moving average.
Given below are the actual sale of a toy for the past 5 weeks. We need to predict the sales for
the 6th week.
W EEK ACTUAL FORECAST CALCULATION
SALES (IN UNITS)
(IN UNITS)
1 1634
2 1821
3 2069
4 1952
5 2178 1869 (1634+1821+2069+ 1952)/4
6 2005 (1821+2069+1952+2178)/4
16. Weighted Moving Average
Method
The data in the recent past periods should be given more weight or
importance compared to the data in the periods far off from the
current time.
W EEK ACTUAL FOR ECAST CALCULATION
SALE (IN
(IN UNITS) UNITS)
1 1634(0.1)
2 1821(0.2)
3 2069(0.3)
4 1952(0.4)
5 1929 (1634*0.1+1821*0.2+20
69*0.3+1952*0.4)/ 1
17. Linear Regression Analysis
It is applied in situations where two variables are
linearly correlated to each other.
In time series analysis, the independent variable
is time while the dependent variable is the actual
demand in the past.
A graph showing the points for the corresponding
values of two variables is called scatter diagram.
These points should display an approximately
linear trend.
18. Example of linear regression
Y= 1060X + 440 is the regression equation
Interpretation: As the age of the car increase by 1 year
the maintenance cost is expected to increase by Rs1060.
19. How to choose a demand forecasting
technique
Objectives of a forecast
Cost involved
Time perspective (short run or long run)
Complexity of the technique
Nature and quality of available data
21. The problem with Moving Averages
Methods
Forecast lags with increasing demand
Forecast leads with decreasing demand
22. Exponential Smoothing
Methods
Single Exponential Smoothing
–– Similar to single Moving Average
Double (Holt’s) Exponential Smoothing
–– Similar to double Moving Average
–– Estimates trend
Triple (Winter’s) Exponential Smoothing
–– Estimates trend and seasonality
24. Holt’s Exponential smoothing
(Double Exponential Smoothing)
Sometimes called exponential smoothing with
trend.
If trend exists, single exponential smoothing
may need adjustment.
There is a need to add a second smoothing
constant to account for trend.
It adds a growth factor (or trend factor) to the
smoothing equation as a way of adjusting for the
trend
25. Winter’s Exponential Smoothing
(Triple Exponential Smoothing)
Winter’s exponential smoothing model is the
second extension of the basic Exponential
smoothing model.
It is used for data that exhibit both trend and
seasonality.
It is a three parameter model that is an extension of
Holt’s method.
An additional equation adjusts the model for the
seasonal component.
26. TREND ANALYSIS
Forecasting method used in causal quantitative
analysis based upon linear regression analysis.
The dependent variable should have a causal
relationship with the independent variable.
For eg.
Dependent variable : No. of units produced
Independent variable : No. of labors present
28. MEASUREMENT OF
FORECASTING ERRORS
Running sum of forecast errors
Mean forecast error
Mean absolute deviation
Mean squared error
Mean absolute percentage error
Tracking signal
29. Tracking signal
Dynamic measure of forecasting errors as can be
updated after every time new actual demand data
is added.
TS=RSFE/MAD
In ideal forecast system ,TS should hover closely
around zero.
Region above centre zero line shows
Actual demand > forecast
Region below centre zero line shows
Actual demand < forecast
31. Forecast Control Limits
Used in controlling the forecasting errors.
Here assumed that forecasting errors follow a
normal distribution curve and are randomly
distributed around the mean(assumed,=0).
Forecasting system is said to be performing well if
all the forecast error points fall within the control
limit.
Upper control limit= 0+3s (s=(MSE)½)
Lower control limit= 0-3s (s=(MSE)½)
Any point not lying in the limit is a signal to
forecaster to look for cause of variation.
Notas do Editor
In this method, the forecast is the average of the last “x” number of observations, where “x” is some suitable number. Suppose a forecaster wants to generate three-period moving averages. In the three-period example, the moving averages method would use the average of the most recent three observations of data in the time series as the forecast for the next period. This forecasted value for the next period, in conjunction with the last two observations of the historical time series, would yield an average that can be used as the forecast for the second period in the future. The calculation of a three-period moving average is illustrated in following table. In calculating moving averages to generate forecasts, the forecaster may experiment with different-length moving averages. The forecaster will choose the length that yields the highest accuracy for the forecasts generated.
The difference between trend analysis and linear regression is that the independent variables can be any other variable except time.
Day 21,22,23 form a straight line which is best fit line
It is always desired that demand forecast value should be as close as possible to the actual demand . But some forecasting errors do take place and we need to measure them so that steps to minimize them can be taken.
Mean forecast error assumed to be =0. The causes may be temporary shortage, natural phenomena such as change in weather conditions, mistake in calculation etc.