Forecasting

What is Forecasting?

 Process of predicting a future event and it is

a mere guess.
 It is the estimating the future demand for
products and services are commonly referred
as a sales forecast
 Underlying basis of all business decisions:
 Production
 Inventory
 Personnel
 Facilities
NEED OF DEMAND FORECASTING

New facility planning

Production planning
Workforce scheduling
Financial planning
Forecasts by Time Horizon

 Short-range forecast
 Up to 1 year (usually less than 3 months)
 Job scheduling, worker assignments, plan
for purchasing
 Medium-range forecast
 3 months to 3 years
 Sales & production planning, budgeting

 Long-range forecast
 3 years, or more
 New product planning, facility location
Types of Forecasts

 Economic forecasts
 Address the future business conditions
(e.g., inflation rate, money supply, etc.)
 Technological forecasts
 Predict the rate of technological progress
 Predict acceptance of new products

 Demand forecasts
 Predict sales of existing products
Features of demand forecasting

 It generally assume the same underlying

reasons
 Forecasts are rarely perfect
 Forecast for group items will be more
perfect than the individual items
 Forecast accuracy decreases as the
time period covered by the forecast
Seven Steps in Forecasting
 Determine the purpose of the forecast
 Select the items to be forecasted
 Determine the time horizon of the
forecast
 Select the forecasting model(s)
 Gather the data
 Make the forecast
 Validate and implement results
Objectives of demand forecasting

Short range objectives

• Formulation of production strategy and
policy
• Formulation of pricing policies
• Planning and control of sales
• Financial planning
Objectives of demand forecasting

Medium or Long range objectives

• Long range planning for production
capacity
• Labour requirements
• Restructuring the capital structure
Forecasting Approaches
Qualitative Methods Quantitative Methods
 Used when situation is  Used when situation
vague & little data is stable & historical
exist data exist
 New products  Existing products
 New technology  Current technology
 Involves intuition,  Involves
experience mathematical
 e.g., forecasting sales on techniques
Internet  e.g., forecasting sales of
color televisions
Qualitative Methods
 Jury of executive opinion
 Pool opinions of high-level executives, sometimes
augment by statistical models
 Delphi method or judge mental method
 Panel of experts, queried iteratively
 Sales force composite
 Estimates from individual salespersons are
reviewed for reasonableness, then aggregated
 Consumer (Market research) Survey
 Ask the customer
Quantitative Approaches

Time series model(Trend, Seasonality, Cycles)

Naive approach
Moving average
Exponential smoothing
Casual models
Trend projection
Linear regression analysis
Time Series Models
 Set of evenly spaced numerical data
 Obtained by observing response variable at
regular time periods
 Forecast based only on past values
 Assumes that factors influencing past and
present will continue influence in future
 Example
Year: 1998 1999 2000 2001 2002
Sales: 78.7 63.5 89.7 93.2 92.1
Time Series Components

Trend Cycle

Seasonal Random
Trend Component
 Persistent, overall upward or downward
pattern
 Due to population, technology etc.
 Several years duration
Seasonal Component

 Regular pattern of up & down

fluctuations
 Due to weather, customs, etc.
 Occurs within 1 year
Cyclical Component
 Repeating up & down movements
 Due to interactions of factors influencing
economy
 Can be anywhere between 2-30+ years
duration
Random Component

 Erratic, unsystematic, ‘residual’ fluctuations

 Due to random variation or unforeseen events
 Union strike
 Tornado

 Short duration & non-repeating

1.Naive Approach

 Assumes demand in next period is equal to

the actual demand in most recent period
 e.g., If May sales were 48, then June sales
will be 48
 Sometimes cost effective & efficient
2.Moving Average Method
 Moving average uses a number of most recent
historical actual data values to generate a forecast.
 MA is a series of arithmetic means
 Used if little or no trend
 Used often for smoothing
 Provides overall impression of data over time
 Equation:

MA = ∑ Demand in Previous n Periods

n
example

 Forecast demand for 4 months

d1+d2+d3 *4
3
3.Exponential Smoothing Method

• It requires only three items of data this

periods forecast, the actual demand for
this period and α which is referred to as
a smoothing constant and having value
between 0 and 1
• Next period’s forecast = This period forecast +
α{this period’s actual dd – this periods
forecast}
Exponential Smoothing
Equations
 Ft = Ft-1 + α(At-1 - Ft-1)
Ft = forecast for this period
Ft-1 = forecast for the previous period
At-1= Actual demand for the previous period
α = σ Smoothing constant (0 to 1)
Linear Trend Projection

 Used for forecasting linear trend line

 Assumes relationship between
response variable, Y, and time, X, is a
linear function
Yi = a + bX i
 Estimated by least squares method
 Minimizes sum of squared errors
Correlation

 Answers: ‘how strong is the linear

relationship between the variables?’
 Coefficient of correlation Sample
correlation coefficient denoted r
 Range: -1 < r < 1
 Measures degree of association

 Used mainly for understanding

Linear regression analysis

 The demand or sales forecast is a

dependent variable and other factors are
independent variables
Factors to be considered in the
selection of forecasting method
 Cost and accuracy
 Data available
 Time span
 Nature of products and services
 Impulse response and noise dampening
Selecting a Forecasting Model

 You want to achieve:

 No pattern or direction in forecast error

^
Error = (Yi - Yi) = (Actual - Forecast)
 Seen in plots of errors over time
 Smallest forecast error
 Mean Absolute Deviation (MAD), or Mean
Absolute Percentage Error (MAPE)
 Mean Squared Error (MSE)
Which Model Is “Best” So Far?

 The Naïve model has both the lowest

MAD (1.91) and MSE (4.45) of the first
five models tested
 Therefore, the Naïve model is the “best”
 However, it may be that one model has
the lowest MAD or MAPE and another
model has the lowest MSE…
So Which Model Do You Choose?

 If you only require the forecast with the

smallest average deviation, choose the
model with the smallest MAD or MAPE
 However, if you have a low tolerance for
large deviations choose the model with
the smallest MSE