FORECASTING

What is Forecasting? Caution: The Realities!

✓ Process of predicting a future event ✓ Forecasts are seldom perfect.
✓ Underlying basis of all business decisions ✓ Most techniques assume an underlying stability
• Production in the system.
• Inventory ✓ Product family and aggregated forecasts are
• Personnel more accurate than individual product
• Facilities forecasts.

Types of Forecasts (based on use) Forecasting Techniques/Approaches

1. Economic forecasts A. Qualitative Forecast/Methods - method of
• Address business cycle – inflation rate, making predictions about a company's finances
money supply, housing starts, etc. that uses judgment from experts.
2. Technological forecasts ✓ Used when situation is vague and little data
• Predict rate of technological progress exist.
• Impacts development of new products • New products
3. Demand forecasts • New technology
• Predict sales of existing products and ✓ Involves intuition, experience.
services • e.g., forecasting sales on Internet

Forecasting Time Horizons (type of forecast based on Types of Qualitative Methods

period covered) 1. Jury of executive opinion
1. Short-range forecast • Pool opinions of high-level experts,
• Up to 1 year, generally less than 3 months sometimes augment by statistical models.
• Purchasing, job scheduling, workforce levels, 2. Delphi method
job assignments, production levels • Panel of experts queried iteratively.
2. Medium range forecast 3. Sales force composite
• 3 months to 3 years • Estimates from individual salespersons are
• Sales and production planning, budgeting reviewed for reasonableness, then
3. Long-range forecast aggregated.
• 3+ years 4. Consumer Market Survey
• New product planning, facility location, • Ask the customer.
research, and development
B. Quantitative Methods - relies on numerical
Distinguishing Differences values and calculations to make predictions and
✓ Medium/long range forecasts deal with more inform decision-making.
comprehensive issues and support management ✓ Used when situation is ‘stable’ and historical
decisions regarding planning and products, data exist.
plants and processes • Existing products
✓ Short-term forecasting usually employs • Current technology
different methodologies than longer-term ✓ Involves mathematical techniques.
forecasting. • e.g., forecasting sales of color televisions
✓ Short-term forecasts tend to be more accurate
than longer-term forecasts. Types of Quantitative Methods
1. Naive approach
Strategic Importance of Forecasting
✓ Human Resources – Hiring, training, laying off
workers.
2. Moving averages
3. Exponential smoothing
4. Trend projection
5. Linear regression
} Time-Series
Models
Associative
✓ Capacity – Capacity shortages can result in
undependable delivery, loss of customers, loss Time Series Forecasting Model
of market share. ✓ Set of evenly spaced numerical data.
✓ Supply Chain Management – Good supplier • Obtained by observing response variable at
relations and price advantages regular time periods.
✓ Forecast based only on past values, no other
Seven Steps in Forecasting variables important.
1. Determine the use of the forecast. • Assumes that factors influencing past, and
2. Select the items to be forecasted. present will continue influence in future.
3. Determine the time horizon of the forecast.
4. Select the forecasting model(s) Time Series Components
5. Gather the data. • Trend
6. Make the forecast. A long-term upward or downward movement in data.
7. Validate and implement results.
 • Seasonality ∑𝑛𝑖=1 𝐴𝑖
𝐹𝑡 =
Short—term regular variations related to the calendar 𝑛
or time of day. where:
• Cycle i = refers to the most recent period
Wavelike variations lasting more than one year. n = number of periods (data points) in the moving
• Irregular variation average
Caused by unusual circumstances, not reflective of Ai = actual value with age I
typical behavior. Ft = forecast
• Random variations
Residual variations after all other behaviors are Sample Problem: Donna’s Garden Supply wants a 3-
accounted for. month moving average forecast, including a forecast for
the next January for shed sales.
Trend Component
✓ Persistent, overall upward, or downward pattern Month Actual Month Actual
✓ Changes due to population, technology, age, Shed Shed
culture, etc. Sales Sales
✓ Typically, several years duration
(units) (units)
Seasonal Component January 10 July 26
✓ Regular pattern of up and down fluctuations
✓ Due to weather, customs, etc. February 12 August 30
✓ Occurs within a single year.
March 13 September 28
Cyclical Component April 16 October 18
✓ Repeating up and down movements
✓ Affected by business cycle, political, and May 19 November 16
economic factors.
✓ Multiple years duration June 23 December 14
✓ Often causal or associative relationships.

Random Component
✓ Erratic, unsystematic, ‘residual’ fluctuations Solution:
✓ Due to random variation or unforeseen events
✓ Short duration and nonrepeating

Naive Approach
✓ Assumes demand in next period is the same as
demand in most recent period.
• e.g., If January sales were 68, then February
sales will be 68.
✓ Sometimes cost effective and efficient.
✓ Can be good starting point.

Sample Problem
National Mixer Inc., sells can openers. Monthly sales for
a seven-month period were as follows:
Weighted Moving Average
Month Feb. Mar. Apr. May Jun. Jul. Aug. ✓ Used when trend is present.
Sales 19 18 15 20 18 22 20 • Older data usually less important
(000 ✓ Weights based on experience and intuition.
units)
𝑭𝒕 = ∑ 𝑾𝒊 𝑨𝒊
Forecast September sales volume using naïve
forecasting. where:
wi = assigned weight for each Ai
Solution & Answer: Ai = actual value for each i
𝐹𝑆𝑒𝑝𝑡𝑒𝑚𝑏𝑒𝑟 = 20,000 𝑢𝑛𝑖𝑡𝑠
Sample Problem: Donna’s Garden Supply (from the
Moving Average Method previous problem) wants to forecast storage shed sales
✓ MA is a series of arithmetic means. by weighting the past 3 months, with more weight given
✓ Used if little or no trend. to recent data to make them more significant.
✓ Used often for smoothing.
• Provides overall impression of data over time.
 National Mixer Inc. sells can openers. Monthly sales for
a seven-month period were as follows:
Month Feb. Mar. Apr. May Jun. Jul. Aug.

Sales 19 18 15 20 18 22 20
(000
units)
Solution:

Forecast September sales volume using Exponential

smoothing with smoothing constant equal to 0.20, and
assuming a March forecast of 19(000)

𝐹𝑆𝑒𝑝𝑡𝑒𝑚𝑏𝑒𝑟 = 19,261.184 ≅ 19,261 𝑢𝑛𝑖𝑡𝑠

Potential Problems with Moving Average
✓ Increasing n smooths the forecast but makes it
less sensitive to changes. Selecting the Smoothing Constant (α)
✓ Do not forecast trends well. • The appropriate value of the smoothing
✓ Require extensive historical data. constant (α) can make a difference between an
Exponential Smoothing accurate forecast and inaccurate forecast.
✓ Sophisticated weighted moving averaging method • High values of α are chosen when the
that is still relatively easy to use and understand. underlying average is likely to change.
✓ Each new forecast is based on the previous forecast • Low values of α are chosen when the underlying
plus a percentage of the difference between that average is fairly stable.
forecast and the actual value of the series at that Notes: Smoothing constant is generally in the range of
point. 0.05 to 0.50 for business applications. It can be changed
➢ Require extensive historical data. to give more weight to recent data (when α is high) or
more weight to past data (when α is low). When α

Ft = Ft - 1 + α(At - 1 -Ft - 1) reaches the extreme 1.0, then it becomes identical to

naïve forecasting model. In picking the value for
smoothing constant, the objective is to obtain the most
where: accurate forecast.
Ft = forecast for period t
Ft-1 = forecast for period t-1 Exponential Smoothing with Trend Adjustment
α = smoothing constant ✓ More sophisticated forecasting method used
At-1 = actual value for period t-1 commonly when there is a trend component in
the time series.
Sample Problem: In January, a car dealer predicted ✓ The idea behind Trend-Adjusted Exponential
February demand for 142 Ford Mustangs. Actual Smoothing for making forecasts consists of
February demand was 153 autos. Using a smoothing using an exponential smoothing form of
constant chosen by the management α =0.20, the forecasting, but with a correction to account for
dealer wants to forecast March demand using a trend (when it exists.
exponential smoothing model. ✓ This technique requires two smoothing
constants: α for the average and β for the trend.
Solution: 𝑭𝒕 = 𝜶(𝑨𝒕−𝟏 ) + (𝟏 − 𝜶)(𝑭𝒕−𝟏 + 𝑻𝒕−𝟏 )
Predicted demand = 142 Ford Mustangs 𝑻𝒕 = 𝜷(𝑭𝒕 − 𝑭𝒕−𝟏 ) + (𝟏 − 𝜷)𝑻𝒕−𝟏
Actual demand = 153 𝑭𝑰𝑻𝒕 = 𝑭𝒕 + 𝑻𝒕
Smoothing constant a = .20 where:
Ft = exponentially smoothed forecast of the data for
Ft = Ft - 1 + α(At - 1 -Ft - 1) period t
𝐹𝑀𝑎𝑟𝑐ℎ = 142 + 0.20(153 − 142) Ft-1 = exponentially smoothed forecast for period t-1
= 144.20 ≅ 144 𝑎𝑢𝑡𝑜𝑠 Tt = exponentially smoothed trend in period t
Tt-1 = exponentially smoothed trend in period t-1
At-1 = actual value for period t-1
α = smoothing constant (0 ≤ α ≤ 1)
β = smoothing constant for trend (0 ≤ β ≤ 1)
Steps to follow:
1. Compute for Ft, the exponentially smoothed
forecast for period t
2. Compute the smoothed trend, Tt
3. Calculate the forecast including trend FIT, by FIT
= Ft + Tt

Sample Problem: A large Portland manufacturer wants

to forecast demand for a piece of pollution-controlled
equipment. A review of the past sales, as shown below,
indicates that an increasing trend is present:

Month Actual Month Actual

Demand Demand
January 12 June 21
February 17 July 31
March 20 August 28
Exponential Smoothing with Trend Adjustment (Graph
April 19 September 36 of the Sample Problem)

May 24 October ?

Smoothing constants are assigned the values of α = 0.20

and β = 0.40. The firm assumes the initial forecast for
the January was 11 units and the trend over that period
was 2 units. Complete the forecast for the 10-month
period.

❖ Trend Projections
✓ Fitting a trend line to historical data points to
project into the medium to long-range.
✓ Linear trends can be found using the least
squares technique.
𝒚 = 𝒂 + 𝒃𝒙
Where:
y = computed value of the variable to be predicted
(dependent variable)
a = y-axis intercept
b = slope of the regression line
x = the independent variable

Least Squares Method

Equations to calculate the regression variables.
𝒏 ∑ 𝒙𝒚 − ∑ 𝒙 ∑ 𝒚
𝒃=
𝒏 ∑ 𝒙𝟐 − (𝒙𝟐 )

∑𝒚 − 𝒃 ∑𝒙
𝒂=
𝒏

Sample Problem: The demand for electric power at N.Y

Edison over the period of 2001 to 2007 is shown in the
table below. The firms want to forecast 2008 demand by
fitting a straight-line trend to these data:
 Solution:
∑ 𝐴𝑣𝑒. 𝐴𝑛𝑛𝑢𝑎𝑙 𝐷𝑒𝑚𝑎𝑛𝑑
𝐴𝑣𝑒. 𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝐷𝑒𝑚𝑎𝑛𝑑 =
𝑛

1128
1. Least Squares Requirements 𝐴𝑣𝑒. 𝑀𝑜𝑛𝑡ℎ𝑙𝑦 𝐷𝑒𝑚𝑎𝑛𝑑 = = 94 𝑢𝑛𝑖𝑡𝑠
12 𝑚𝑜𝑛𝑡ℎ𝑠
1. Always plot the data to insure a linear
relationship. To get the Seasonal Index:
2. Do not predict time periods far beyond the
database.
3. Deviations around the least squares line are
assumed to be random.

Seasonal Variations in Data

• Regular up-and-down movement in a time
series that relate to recurring events such as
weather or holidays.
• Seasonality may be applied to hourly, daily,
weekly, monthly, or other recurring patterns.
• Understanding seasonal variations is important
for capacity planning in organizations that
handle peak loads.
• Expressed in terms of the amount that actual
values differ from average values in the time
series.

Multiplicative Seasonal Model

• Seasonal factors are multiplied by an estimate of
average demand to produce seasonal forecast.
Steps in the process:
1. Find average historical demand for each season.
2. Compute the average demand over all seasons.
3. Compute a seasonal index for each season.
4. Estimate next year’s total demand.
5. Divide this estimate of total demand by the
number of seasons, then multiply it by the
seasonal index for that season.

Sample Problem ADM Company, a distributor of Sony

laptop wants to develop monthly indices for sales. Data
from 2005-2007 monthly sales are shown in the table.
Compute for the seasonal indices.
To get Forecasting: Solution:

Associative Forecasting
• Used when changes in one or more independent
variables can be used to predict the changes in
the dependent variable.
• Uses linear regression analysis to make prediction.
𝒚 = 𝒂 + 𝒃𝒙

𝒏 ∑ 𝒙𝒚 − ∑ 𝒙 ∑ 𝒚 2. If the local chamber of commerce predicts that

𝒃= the West Bloomfield area payroll will be $6B
𝒏 ∑ 𝒙𝟐 − (𝒙𝟐 )
next year, what is the probable estimate of sales
∑𝒚 − 𝒃 ∑𝒙 for Nodel Construction Company.
𝒂= 𝒚 = 𝟏. 𝟕𝟓 + 𝟎. 𝟐𝟓𝒙
𝒏 𝒚 = 𝟏. 𝟕𝟓 + 𝟎. 𝟐𝟓($𝟔𝑩) = $𝟏, 𝟓𝟎𝟎, 𝟎𝟎𝟎, 𝟎𝟎𝟐
Where:
Correlation
y = computed value of the variable to be
• How strong is the linear relationship between
predicted (dependent variable)
the variables?
a = y-axis intercept
• Correlation does not necessarily imply causality!
b = slope of the regression line
• Coefficient of correlation, r, measures degree of
x = the independent variable though to predict the value
association
of the dependent variable
• Values range from -1 to +1
Sample Problem:
PEARSON PRODUCT-MOMENT CORRELATION
1. Nodel Construction Company renovates old
COEFFICIENT
homes in West Bloomfield, Michigan. Over time,
𝒏 ∑ 𝒙𝒚 − ∑ 𝒙 ∑ 𝒚
the company has found that its dollar volume of 𝒓=
renovation work is dependent on the West √[𝒏 ∑ 𝒙𝟐 − (∑ 𝒙)𝟐 ] [ 𝒏 ∑ 𝒚𝟐 − (𝒙)𝟐
Bloomfield area payroll. Management wants to
establish a mathematical relationship to help
predict sales.
 Multiple Regression Analysis
• If more than one independent variable is to be
used in the model, linear regression can be
extended to multiple regression to
accommodate several independent variables.

Where:
y = dependent variable
a = constant, the y intercept
𝒙𝟏 , 𝒙𝟐 = values of the two independent variables
𝒃𝟏 ,𝟐 = coefficients for the two independent variables

Note: The mathematics of multiple regression becomes

quite complex as numerous independent variables are
used in making prediction, hence, this is usually tackled
by computer