Forecasting

What is forecasting all about?

Demand for a product We try to predict the future by

looking back at the past

Predicted
demand
looking back
Time six months
Jan Feb Mar Apr May Jun Jul Aug

Actual demand (past sales)

Predicted demand
 Why forecasting requires?

• Consider the example of Dell.

• Dell performs assembly of P.C. components in
response to the customer orders
• To determine the amount of components to have
on hand and to determine the capacity needed in
its plants
• Dell requires the forecast of future demand.
 Characteristics of forecasting
• A causal relationship is needed for forecasting.
• If what happens is purely random and does not depend on
anything, you cannot predict what will happen.
• If a student got A grades only during his/her first two
years, the probability that he/she fails an exam is smaller.

Forecasting
Previous data Future data
 Forecasting Time Horizons
Ø Short-range forecast
þ Forecasting will be done up to 3 months.
þ Purchasing, workforce levels, job assignments
Ø Medium-range forecast
þ 3 months to 3 years
þ Sales and production planning, budgeting
Ø Long-range forecast
þ 3+ years
þ New product planning, facility location, research and
development
 Types of Forecasts

– Economic forecasts
– Predict a variety of economic indicators like inflation
rates, interest rates, etc.

Technological forecasts

– Predict rates of technological progress and innovation.

Demand forecasts

– Predict the future demand for a company’s products.

Type of Forecasting Methods

Qualitative Quantitative
methods methods
 Qualitative methods

• These types of forecasting methods are based on

judgments, opinions, intuition, emotions, or
personal experiences of expert.
• It is subjective in nature.
• They do not rely on any mathematical
computations.
 Quantitative methods

• These types of forecasting methods are based on

mathematical (quantitative) models.
• It is objective in nature.
• They rely heavily on mathematical computations.
 Qualitative Forecasting Methods

Jury of Executive Delphi Sales Force Market

Opinion Method Composite Survey
 Jury of Executive Opinion
• Approach in which a group of managers meet
with their own idea about the forecast.
• They discuss, revise their opinions according to
other’s opinions
• Finally they develop a forecast.
• Hardware accessories manufacturer uses this
method to estimates order.
 Delphi Method
• Approach in which consensus agreement is
reached among a group of experts based on
questionnaires.
• Automobile manufacturer uses this method
 Sales Force Composite
• Approach in which each salesperson estimates
sales in his or her region.
• Then estimated sales are combined at district
level followed by state level and then nation
level.
• Ex: Insurance Company about insurance
policies, Medicare supplement policies.
 Market Survey
• Approach that uses interviews, feedback form
and surveys to judge preferences of customer
and to assess demand.
• It is purely based on customer opinions.
• Apple use market survey to find out what
exactly customer wants from their device.
• They then figure out how to make those wants
in a reality.
 Quantitative Forecasting Methods

Time-Series Models Associative Models

 Time-Series Models
• Time series models look at past patterns of data
and attempt to predict the future based upon the
underlying patterns contained within those data.
• Ex: If we are predicting the demand of Umbrella
then we have to use the past sales of umbrella to
make the forecasts
 Associative Models
• Associative models incorporate the variables
that might influence the quantity being
forecast.
• For example, an associative model for
Umbrella sales might use factors such as
weather, advertising budget, and competitors’
prices.
 Quantitative Forecasting Methods

Quantitative
Forecasting

Time Series Associative

Models Models

Naive Simple Moving Exponential Trend

Approach Mean Average Smoothing Projection

Linear Multiple
Regression Regression
 Naive Approach
þ Assumes demand in next period is the same
as demand in most recent period
þ e.g., In other words, if sales of a product, say
MI cell phones were 68 units in September,
þ Then we can forecast that October sales will
also be 68 phones.
• Where
• Ft+1 = forecast for next period, t+1
• At = Actual value for current period, t
• t = current time period
 Example: 1

• A restaurant is forecasting sales of chicken

dinners for the month of April. Total sales
of chicken dinners for March were 320. If
management uses the naive method to
forecast, what is their forecast of chicken
dinners for the month of April?
 Simple Mean Method

• One of the simplest forecasting models is the

simple mean or average.
• Here the forecast is made by simply taking an
average of all past sales data.
• Where
• Ft+1 = forecast for next period, t+1
• At = Actual value for current period, t
• t = current time period
• n = No. of past data available
 Example: 2
• New Tools Corporation is forecasting sales for its
classic product, Handy-Wrench. Handy-Wrench
sales have been steady, and the company uses a
simple mean to forecast. Weekly sales over the
past five weeks are available. Use the mean to
make a forecast for week 6.
 Moving Average Method

• A moving-average forecast uses a specific

number of historical actual data values to
generate a forecast.
• We consider only latest data for forecasting.
• Moving averages are useful if market demands
will stay fairly steady over time.
Where ‘n’ is number of latest period involved in
moving average
 Example: 3
• Actual sales data of product for a 6 months are given in
following Table. Make a forecast for the month of July
using three period moving average and five period
moving average.
Months Actual Sales
January 260
February 245
March 210
April 200
May 250
June 275
July
Example
 Weighted Moving Average
þ The forecast for next period (t+1) will be equal
to a weighted average of a specified number of
the most recent observations.
þ Weights based on experience and decision of
expert
 Example: 4
• A manager at Fit Well department store wants to forecast
sales of swimsuits for August using a three-period
weighted moving average. Sales for May, June, and July
are as follows:

• The manager has decided to assign more weight to recent

data. The weights for swimsuits are applied as 3,2,1.
 Example: 5
• Shipments (in tons) of welded tube by an aluminium
producer are shown below:

a) Plot the data on graph and comment on the relationship.

b) Using 3-year moving average forecast shipments for
year 4 to year 12.
c) Using a weight of 3 for the most recent data, 2 for the
next, and 1 for the oldest, forecast shipments for year 4
to 12.
 Exponential Smoothing
Ø The new forecast for next period (t) will be
calculated as follows:

New forecast = Last period’s forecast + a (Last period’s actual

demand – Last period’s forecast)

Ft = Ft-1 + a(At-1 – Ft-1)

Where
Ft = new forecast
Ft-1 = previous period’s forecast
a = smoothing coefficient whose (0≤ a ≤ 1)
At-1 = previous period’s actual demand

ØThe smoothing constant, α, is generally in the

range from .05 to .50 for business applications.
ØWhen α reaches the extreme of 1.0, then
exponential smoothing becomes as Neive
approach.
 Example: 6
• A firm uses simple exponential smoothing (α
= 0.1) to forecast demand. The forecast for the
week 1 was 500 units, whereas actual demand
turned out to be 450 units. Forecast the demand
for the week 2.
Example: 7
 Forecast Error
• A forecasts will always deviate from the actual
demand.
• A large degree of error may indicate that the
forecasting technique is the wrong, one has to
change the technique.
• Forecast error is the difference between the actual
and the predicted value of a time series data.
• Forecast error can be given by.

Forecast error = Actual demand - Forecast value

Where
• et = forecast error at period t
• At = Actual demand at period t
• Ft = Forecast value at period t
 Forecast error measures

• There are three measures used in practice to

calculate the overall forecast error.
o Mean Absolute Deviation (MAD)
o Mean Squared Error (MSE)
o Mean Absolute Percent Error (MAPE)
 Mean Absolute Deviation (MAD)

• This value is computed by taking the sum of the

absolute values of the individual forecast errors
(deviations) and dividing by the number of
periods of data (n)
 Example: 8
• During the past 8 quarters, the Port of Baltimore has
unloaded large quantities of grain from ships. The port’s
operations manager wants to test the use of exponential
smoothing to see how well the technique works in
predicting tonnage unloaded. He guesses that the forecast
of grain unloaded in the first quarter was 175 tons. The
values of α is be examined is 0.10. Compare the actual
data with the data we forecast and then find the mean
absolute deviation.
Quarter Actual Tonnage Unloaded Quarter Actual Tonnage Unloaded
1 180 5 190
2 168 6 205
3 159 7 180
4 175 8 182
 Example: 9
• A company is comparing the accuracy of two
different forecasting methods. Use MAD to
compare the accuracies of these methods for
the past five weeks of sales. Which method
provides greater forecast accuracy?
Week Actual Forecast with Forecast with
Sales Method A Method A
1 25 30 30
2 18 20 16
3 26 23 25
4 28 29 30
5 30 25 25
 Mean Squared Error (MSE)

• The mean squared error (MSE) is a second way

of measuring overall forecast error.
• MSE is the average of the squared differences
between the actual and forecasted values.
 Example: 10
• Sales of Volkswagen’s popular Beetle have
grown steadily at auto dealerships in Nevada
during the past 5 years (see table below). The
sales manager had predicted before the new
model was introduced that first year sales would
be 410 VWs. Using exponential smoothing
constant α = .30, develop forecasts for years 2
through 5, and identify MSE
 Example:11

• Anandi Beach Resort is a popular resort at Goa.

Table gives the details of actual demand (in
units), which represents the number of
registrations of customers the resort receives in
past 15 days. If the forecast of demand at day
one is 700 units apply exponential smoothing
forecasting method to arrive at forecast for 15
days taking α = 0.5. Measure the overall forecast
error by applying Mean Square Error.
Day Actual Demand (in units)
1 721
2 801
3 854
4 826
5 802
6 897
7 969
8 1078
9 1192
10 1064
11 1005
12 1275
13 1392
14 1458
15 1503
 Mean Absolute Percent Error (MAPE)

• A problem with both the MAD and MSE is that

their values depend on the magnitude of the item
being forecast.
• If the actual and forecast demand is in figure of
thousands, the MAD and MSE values can be
very large.
• To avoid this problem, we can use the Mean
Absolute Percent Error (MAPE) .
 Example: 12

• Table gives the data for actual demand of a

product D and forecast F for a product for last 12
months. Using Mean Absolute Percentage Error
(MAPE) measure forecast error?
Month Actual Demand Forecasts Month Actual Demand Forecasts
1 1578 1584 7 2001 1982
2 1689 1699 8 2053 2073
3 1795 1724 9 1856 1866
4 1522 1512 10 1685 1673
5 1643 1687 11 1715 1720
6 1894 1901 12 1811 1801
 Example: 13

• Bhavishya Life Insurance Company has its head

quarters based at Chennai. Table gives the details
of the actual number of insurance policies
subscribed by it throughout the country (in units)
in past 2 weeks (14 days). Apply exponential
smoothing to arrive at the forecasts for given 14
days by taking α = 0.5 and Make graph to
compare the actual demand pattern with the
forecast and also identify MAPE. (Consider
forecast demand for day one is 120 units)
Day Actual Demand Day Actual Demand
(Units) (Units)
1 125 8 172
2 138 9 160
3 149 10 179
4 130 11 184
5 145 12 197
6 158 13 203
7 166 14 213
Comparison of Measures of Forecast Errors
 Trend Projections
• In this method a straight line is drawn through
the historical data points in a fashion that all
points should be as close as possible to the
straight line .
• For example consider actual demand for a
product in past 1 week (7 days) is as follows:
Day Actual Demand Day Actual Demand
1 125 5 145
2 130 6 158
3 149 7 166
4 130
• Now to obtain forecasted demand at any point of
time, go on straight line on that point of time,
project point on straight line to y axis.
How to draw straight line
• The equation of line is:

• Where b = slope of line

a = point of intersection at y axis
Slope: Slope of line is the measure of the
steepness of line.
 a and b can be given by following equation

or
 Example: 14
• The demand for electric power at N.Y. Edison
over the past 7 years is shown in the following
table, in megawatts. The firm wants to forecast
next year’s demand by fitting a straight-line trend
to these data.
 Example: 15
• Aroma Drip Coffee Incorporation produces
commercial coffee machines that are sold all
over the world. The company’s production
facility has operated at near capacity for over a
year now. Plant Manager, thinks that sales
growth will continue and he want to develop long
range forecasts to help to plan facility
requirements for next 3 years. Sales records for
the past 10 years have been complied. Forecast
the demand for next 3 years using Trend
Projection method.
Year Annual Sales (Thousands of Units)
1 1000
2 1300
3 1800
4 2000
5 2000
6 2000
7 2200
8 2600
9 2900
10 3200
 Associative Models
• Associative forecasting models usually consider
several variables that may affect the quantity
being predicted.
• Once these related variables have been found, a
statistical model is built and used to forecast the
item of interest.
• This approach is more powerful than the time-
series methods that use only the historical values
for the forecast variable.
• For example, the sales of Umbrella might get
affected by the factor such as
Øweather,
Øadvertising budget, and
Øcompetitors’ prices.
• The sales of Dell PCs may be related to Dell’s
Øadvertising budget,
Øcompany’s prices,
Øcompetitors’ prices
Øpromotional strategies, and
Øunemployment rates.
• In associative model there are two type of
variables:
1. Independent variable
2. Dependent variable
• Independent variable: There are many factors,
which may affect the forecasting of any product
demand.
• Dependent variable: The dependent variable (sales
of product) is dependent on the independent
variable.
• For Example: In the case of PC:
• Demand would be called the dependent variable.
• The other variables like
ØDell’s advertising budget,
ØThe company’s prices,
ØCompetitors’ prices and
ØPromotional strategies
would be called independent variables .
 Linear Regression
• Linear regression is linear approach to modelling
the relationship between a dependent variable and
one independent variable.

• y = dependent variable
• b = slope of regression line
• a = y-axis intercept
• x = independent variable
 Difference between trend projection and linear
regression
• Trend projection always have positive slope but
linear regression can also have a negative slope
• In trend projection the independent variable is
time where as in linear regression the
independent variable need not be time, can be
any variable.
 Example: 16
• A maker of personalized golf shirts has been
tracking the relationship between sales and
advertising dollars over the past four years. The
results are as follows:

• Use linear regression to find out what sales would be

if the company invested $53,000 in advertising for
next year.
 Example: 17
• The general manager of a building materials
production plant feels that the demand for
plasterboard shipments may be related to the
number of construction permits issued in the
county during the previous quarter. The manager
has collected the data shown in Table
(a) Compute values for the slope b and intercept a.
(b) Determine a point estimate for plasterboard
shipments when the number of construction permits
is 30.
 Multiple Regression
• Multiple regression is approach to modelling the
relationship between a dependent variable and more
than one independent variable.

• y = dependent variable
• a = y-axis intercept
• x1 and x2= values of the two independent variables
• b1 and b2= coefficients for the two independent
variables
• Where
• y = dependent variable
• a = y-axis intercept
• x1, x2 ... xn= values of the n independent variables
• b1, b2 ... bn= coefficients for the n independent
variables