Types of Forecasting Models

 Qualitative methods – judgmental methods

– Forecasts generated subjectively by the
Forecasting forecaster
– Educated guesses
Principles of Forecasting
 Quantitative methods – based on
Many types of forecasting models that differ
mathematical modeling:
in complexity and amount of data & way they
– Forecasts generated through
generate forecasts:
mathematical modeling

 Forecasts are rarely perfect Qualitative Methods

Type Characteristics Strengths Weaknesses


Executive A group of managers Good for strategic or One person's opinion
opinion meet & come up with new-product can dominate the
Forecasts are more accurate for grouped data a forecast forecasting forecast

than for individual items Market Uses surveys & Good determinant of It can be difficult to
research interviews to identify customer preferences develop a good

 Forecast are more accurate for shorter than

Delphi
customer preferences

Seeks to develop a Excellent for

questionnaire

Time consuming to
method consensus among a forecasting long-term develop
longer time periods group of experts product demand,
technological
changes, and
Types of Forecasting Methods
 Decide what needs to be forecast Quantitative Methods
– Level of detail, units of analysis & time  Time Series Models:
horizon required – Assumes information needed to
 Evaluate and analyze appropriate data generate a forecast is contained in a
– Identify needed data & whether it’s time series of data
available – Assumes the future will follow same
 Select and test the forecasting model patterns as the past
– Cost, ease of use & accuracy  Causal Models or Associative Models
 Generate the forecast – Explores cause-and-effect relationships
 Monitor forecast accuracy over time – Uses leading indicators to predict the
future

Types of Forecasting Methods – Housing starts and appliance sales

• Forecasting methods are classified into two
groups:
Time Series Models
• Forecaster looks for data patterns as
– Data = historic pattern + random
variation
• Historic pattern to be forecasted:
– Level (long-term average) – data
fluctuates around a constant mean
 – Trend – data exhibits an increasing or
decreasing pattern Ft 1   C t A t
• Weighted Moving Average:
– Seasonality – any pattern that regularly
repeats itself and is of a constant – All weights must add to 100% or 1.00
length – e.g. Ct .5, Ct-1 .3, Ct-2 .2 (weights add to
– Cycle – patterns created by economic 1.0)
fluctuations – Allows emphasizing one period over
others; above indicates more weight on
• Random Variation cannot be predicted recent data (Ct=.5)
– Differs from the simple moving average
Time Series Patterns that weighs all periods equally - more
responsive to trends

 Exponential Smoothing: Ft 1  αA t   1  α  Ft

Most frequently used time series method

because of ease of use and minimal amount
of data needed

 Need just three pieces of data to start:

– Last period’s forecast (Ft)
– Last periods actual value (At)
Time Series Models – Select value of smoothing coefficient,
 Naive: Ft 1  At between 0 and 1.0


– The forecast is equal to the actual value
observed during the last period – good If no last period forecast is available, average
for level patterns the last few periods or use naive method
 Simple Mean: Ft 1   A t / n

– The average of all available data - good

 Higher values (e.g. .7 or .8) may place too
much weight on last period’s random
for level patterns
variation
Ft 1   A t / n
 Moving Average:
– The average value over a set time Time Series Problem
period • Determine forecast for periods 7 & 8
(e.g.: the last four weeks) • 2-period moving average
– Each new forecast drops the oldest • 4-period moving average
data point & adds a new observation • 2-period weighted moving average with t-1
– More responsive to a trend but still lags weighted 0.6 and t-2 weighted 0.4
behind actual data • Exponential smoothing with alpha=0.2 and
the period 6 forecast being 375
  Smooth the level of the series:
S July  αA t  (1  α)(S t 1  Tt 1 )   0.2 62   0.8 57  15  70

 Smooth the trend:

TJuly  β(S t  St 1 )  (1  β)Tt 1   0.1 70  57    0.915  14.8

Period Actual
1 300
2 315  Forecast including trend:
3 290 FITAugust  S t  Tt  70  14.8  84.8 gallons
4 345
5 320 Linear Trend Line
6 360 A time series technique that computes a forecast
with trend by drawing a straight line through
7 375
a set of data using this formula:
8 Y = a + bx where
Y = forecast for period X
Time Series Problem Solution X = the number of time periods from X = 0
A = value of y at X = 0 (Y intercept)
B = slope of the line

Forecasting Trend
• Basic forecasting models for trends
compensate for the lagging that would
otherwise occur
• One model, trend-adjusted exponential
smoothing uses a three step process
– Step 1 - Smoothing the level of the
Forecasting trend problem: a company uses series
exponential smoothing with trend to forecast usage S t  αA t  (1  α)(S t 1  Tt 1 )
of its lawn care products. At the end of July the
– Step 2 – Smoothing the trend
company wishes to forecast sales for August. July
demand was 62. The trend through June has been Tt  β(S t  S t 1 )  (1  β)Tt 1
15 additional gallons of product sold per month. – Forecast including the trend
Average sales have been 57 gallons per month. The
company uses alpha+0.2 and beta +0.10. Forecast
FITt 1  S t  Tt
for August.
Forecasting Seasonality
 Calculate the average demand per season
 
– E.g.: average quarterly demand
 Calculate a seasonal index for each season of Causal models establish a cause-and-effect
each year: relationship between independent and
– Divide the actual demand of each dependent variables
season by the average demand per
season for that year
 A common tool of causal modeling is linear
regression:

 Additional related variables may require

 Average the indexes by season multiple regression modeling Y  a  bx

– E.g.: take the average of all Spring
indexes, then of all Summer indexes, ... Linear Regression
 Forecast demand for the next year & divide • Identify dependent (y) and independent (x)
by the number of seasons variables
– Use regular forecasting method & • Solve for the slope of the line
divide by four for average quarterly • Solve for the y intercept
demand
b
 XY  n XY

 X  nX
Multiply next year’s average seasonal demand 2 2

by each average seasonal index

• Develop your equation for the trend line
– Result is a forecast of demand for each
Y=a + bX
season of next year

Seasonality problem: a university must develop

forecasts for the next year’s quarterly enrollments.
It has collected quarterly enrollments for the past
two years. It has also forecast total enrollment for
next year to be 90,000 students. What is the
forecast for each quarter of next year?

Linear Regression Problem: A maker of golf

shirts has been tracking the relationship between
sales and advertising dollars. Use linear regression
to find out what sales might be if the company
invested $53,000 in advertising next year.

Causal Models
 Often, leading indicators can help to predict
changes in future demand e.g. housing starts
 Coefficient of determination ( r 2 ) measures
the amount of variation in the dependent variable
about its mean that is explained by the regression
line. Values of ( r 2 ) close to 1.0 are desirable.

Multiple Regression
• An extension of linear regression but:
– Multiple regression develops a
relationship between a dependent
variable and multiple independent
variables. The general formula is:

Measuring Forecast Error

 Forecasts are never perfect
 Need to know how much we should rely on
our chosen forecasting method
 Measuring forecast error:

28202  4 47.25  147.25  E t  A t  Ft

b  1.15
9253  4 47.25  
2
Note that over-forecasts = negative errors and

a  Y  b X  147.25  1.15 47.25 

under-forecasts = positive errors

a  92.9
Measuring Forecasting Accuracy
Y  a  bX  92.9  1.15X
Y  92.9  1.15 53   153.85  Mean Absolute Deviation (MAD)
– measures the total error in a forecast
Correlation Coefficient without regard to sign
How Good is the Fit? MAD 
 actual  forecast
• Correlation coefficient (r) measures the n

direction and strength of the linear

relationship between two variables. The
 Cumulative Forecast Error (CFE)

closer the r value is to 1.0 the better the – Measures any bias in the forecast
regression line fits the data points. CFE    actual  forecast 

r
n   XY    X   Y   Mean Square Error (MSE)
 X     X  Y    Y
2 2
2 2
n * n – Penalizes larger errors
4 28,202   189 589 
  actual - forecast 
2
r  .982
4 87,165    589  MSE 
2 2
4(9253) - (189) *
n
  .982 
2
r 2
 .964
 Tracking Signal
Forecasting Software
– Measures if your model is working  Spreadsheets
CFE – Microsoft Excel, Quattro Pro, Lotus 1-2-3
TS 
MAD – Limited statistical analysis of forecast
data
Accuracy & Tracking Signal Problem: A company is  Statistical packages
comparing the accuracy of two forecasting – SPSS, SAS, NCSS, Minitab
methods. Forecasts using both methods are shown – Forecasting plus statistical and graphics
below along with the actual values for January  Specialty forecasting packages
through May. The company also uses a tracking
signal with ±4 limits to decide when a forecast –Forecast Master, Forecast Pro, Autobox,
should be reviewed. Which forecasting method is SCA
best?
Guidelines for Selecting Software
 Does the package have the features you want?

 What platform is the package available for?

 How easy is the package to learn and use?

 Is it possible to implement new methods?

 Do you require interactive or repetitive

forecasting?

 Do you have any large data sets?

Selecting the Right Forecasting Model

 Is there local support and training available?
1. The amount & type of available data
 Some methods require more data than  Does the package give the right answers?
others
2. Degree of accuracy required Other Forecasting Methods
 Increasing accuracy means more data • Focus Forecasting
3. Length of forecast horizon – Developed by Bernie Smith
 Different models for 3 month vs. 10 – Relies on the use of simple rules
years – Test rules on past data and evaluate
4. Presence of data patterns how they perform
 Lagging will occur when a forecasting • Combining Forecasts
model meant for a level pattern is – Combining two or more forecasting
applied with a trend methods can improve accuracy
  Forecast decisions serve as the basis for
Collaborative Planning Fore-casting & tactical planning; developing worker
Replenishment (CPFR) schedules (Ch 11).
– Establish collaborative relationships Virtually all operations management decisions are
between buyers and sellers based on a forecast of the future.
– Create a joint business plan
– Create a sales forecast Forecasting Across the Organization
– Identify exceptions for sales forecast  Forecasting is critical to management of all
– Resolve/collaborate on exception items organizational functional areas
– Create order forecast – Marketing relies on forecasting to
– Identify exceptions for order forecast predict demand and future sales
– Resolve/collaborate on exception items – Finance forecasts stock prices, financial
– Generate order performance, capital investment
needs..
– Information systems provides ability to
Forecasting within OM: How it all fits share databases and information
together
Forecasts impact not only other business – Human resources forecasts future
functions but all other operations decisions. hiring requirements
Operations managers make many forecasts,
such as the expected demand for a company’s Chapter 8 Highlights
products. These forecasts are then used to • Three basic principles of forecasting are:
determine: forecasts are rarely perfect, are more accurate
 product designs that are expected to sell (Ch 2), for groups than individual items, and are
 the quantity of product to produce (Chs 5 and more accurate in the shorter term than longer
6), time horizons.

 the amount of needed supplies and materials • The forecasting process involves five steps:
(Ch 12). decide what to forecast, evaluate and analyze
appropriate data, select and test model,
Also, a company uses forecasts to generate forecast, and monitor accuracy.
• determine future space requirements (Ch 10),
• capacity and
• Forecasting methods can be classified into

• location needs (Ch 9), and two groups: qualitative and quantitative.
• the amount of labor needed (Ch 11). Qualitative methods are based on the
subjective opinion of the forecaster and
Forecasts drive strategic operations decisions, such quantitative methods are based on
as: mathematical modeling.
 choice of competitive priorities, changes in • Time series models are based on the
processes, and large technology purchases assumption that all information needed is
(Ch 3). contained in the time series of data. Causal
models assume that the variable being
 forecast is related to other variables in the
environment.
• There are four basic patterns of data: level or
horizontal, trend, seasonality, and cycles. In
addition, data usually contain random
variation. Some forecast models used to
forecast the level of a time series are: naïve,
simple mean, simple moving average,
weighted moving average, and exponential
smoothing. Separate models are used to
forecast trends and seasonality.
• A simple causal model is linear regression in
which a straight-line relationship is modeled
between the variable we are forecasting and
another variable in the environment. The
correlation is used to measure the strength of
the linear relationship between these two
variables.
• Highlights con’t

 Three useful measures of forecast error are

mean absolute deviation (MAD), mean square
error (MSE) and tracking signal.

 There are four factors to consider when

selecting a model: amount and type of data
available, degree of accuracy required, length
of forecast horizon, and patterns present in
the data.