Operations

Management
and TQM
O U R L A D Y O F F AT I M A U N I V E R S I T Y

M O N D AY 1 0 : 2 0 A M – 1 : 2 0 P M

2 ND Y R . B S A A N D B S A I S
O perations

M
anagement
2
Forecasting
➢ A key operations planning activity at Walt Disney World Resort.
Analyst must forecast the attendance at each theme park, and this
influences park hours and other strategic plans. Forecasts are also
generated for more specific activities such as guest arrivals at Disney
hotels, park entry turnstiles, restaurants, and merchandise locations.
All these forecasts are incorporated into a labor demand planning
system that matches staffing schedules with anticipated demand.
Forecasting
➢ Forecasting is the process of projecting the values of one or more variables
into the future.
➢ Good forecasts are needed in all organizations to drive analysis and
decisions related to operations. Forecasting is a key component in many
types of integrated operating systems, such as supply chain management,
customer relationship management, and revenue management systems.
➢ Poor forecasting can result in poor inventory and staffing decisions,
resulting in part shortages, inadequate customer service, and many customer
complaints.
Basic Concepts in Forecasting
a) Forecast Planning Horizon
b) Data Patterns in Time Series
c) Forecast Errors and Accuracy
 Forecast Planning Horizon
➢ Forecasts of future demand are needed at all levels of organizational decision-making. The
planning horizon is the length of time on which a forecast is based.
➢ Long-range forecasts cover a planning horizon of 1 to 10 years and are necessary to plan for the
expansion of facilities and to determine future needs for land, labor, and equipment.
➢ Intermediate-range forecasts over a 3-to-12 month period are needed to plan workforce levels,
allocate budgets among divisions, schedule jobs and resources, and establish purchasing plans.
➢ Short-range forecasts focus on the planning horizon of up to three months and are used by
operations managers to plan production schedules and assign workers to jobs, determine short-
term capacity requirements, and aid shipping departments in planning transportation needs and
establishing delivery schedules.
Time Bucket
➢ The unit of measure for the time period used in a forecast.
➢ A time bucket might be a year, quarter, month, week, day, hour, or even a
minute.
➢ For a long-term planning horizon, a firm might forecast in yearly time
buckets; for a short-range planning horizon, the time bucket might be an hour
or less. Customer call centers, for example, forecast customer demand in 5-,
6-, or 10-minute intervals. Selecting the right planning horizon length and
time bucket size for the right situation is an important part of forecasting.
Data Patterns in Time Series
➢ Statistical methods of forecasting are based on the analysis of historical data, called a
time-series.
➢ A time series is a set of observations measured at successive points in time or over
successive periods of time.
➢ A time series provides the data for understanding how the variable that we wish to
forecast has changed historically. For example, the daily ending Dow Jones stock index
is one example of a time series; another is the monthly volume of sales for a product.
➢ To explain the pattern of data in a time series, it is often helpful to think in terms of
five characteristics: trend, seasonal, cyclical, random variation, and irregular (one-time)
variation.
Trend
➢ The underlying pattern of growth or decline in a time series.
➢ A trend shows gradual shifts or movements to relatively higher or lower
values over a longer period of time. This gradual shifting over time is usually
due to such long-term factors as changes in performance, technology,
productivity, population, demographic characteristics, and customer
preferences.
➢ Trends can be increasing or decreasing and can be linear or non-linear.
Seasonal Patterns
➢ Characterized by repeatable periods of ups and downs over short
periods of time.
➢ Seasonal patterns may occur over a year; for example, the demand for
cold beverages is low during the winter, begins to rise during the spring,
peaks during the summer months, and then begins to decline in the
autumn. Manufacturers of coats and jackets, however, expect the opposite
yearly pattern.
Cyclical Patterns
➢ Regular patterns in a data series that take place over long periods of
time. A common example of a cyclical pattern is the movement of stock
market values during “bull” and “bear” market cycles.
Random Variation
➢ Random variation is sometimes called “noise”.
➢ The unexplained deviation of a time series from a predictable pattern
such as a trend, seasonal, or cyclical pattern.
➢ Random variation is caused by short-term, unanticipated, and
nonrecurring factors and is unpredictable. Because of random
variation, forecasts are never 100% accurate.
Irregular Variation
➢ A one-time variation that is explainable. For example, a hurricane can
cause a surge in demand for building materials, food and water.
➢ After the 9/11 terrorist attacks on the United States, many forecasts
that predicted U.S. financial trends and airline passenger volumes had
to be discarded due to the effects of this one-time event.
Forecast Errors and Accuracy
➢ All forecasts are subject to error, and understanding the nature and size of
errors is important to making good decisions.
➢ Denotation of historical values of a time series is by A1, A2,……., At.
➢ In general, At represents the value of the time series for period t.
➢ Let Ft represent the forecast value for period t.
➢ When we make this forecast, we will not know the actual value of the time
series in period t, At. However, once At becomes known, we can assess how
well our forecast was able to predict the actual value of the time series.
Table 1.1
Forecast Error
➢ The difference between the observed value of the time series and the
forecast, or At – Ft. Suppose that a forecasting method provided the forecasts
in column E of Table 1.1 for the call volume time series . The forecast errors
are computed in column F.
Mean Square Error (MSE)
➢ Calculated by squaring the individual forecast errors and then averaging the results over all [ T ]
periods of data in the time series.

∑( A t - Ft )
2

MSE
T
Table 1.1
Mean Absolute Deviation (MAD)
➢ The average distance of all of the elements in a data set from the mean of the same data set.

➢ Mean absolute deviation (MAD) is computed using the formula below:

∑( A t - Ft)
MAD
T
Table 1.2
Mean Absolute Percentage Error
(MAPE)
➢ The average of the percentage error for each forecast value in the time series.

➢ It measures the average magnitude of error produced by a model, or how far off predictions are
on average. Example: If the MAPE value is 20%, it means that the average absolute percentage
difference between the predictions and the actuals is 20%.

MAPE
∑[ ( A t - Ft )
At X 100 ]
T
Table 1.3
MSE VS. MAD
➢ The major difference between MSE and MAD is that MSE is influenced much more
by large forecast errors than by small errors (because the errors are squared).
➢ The values of MAD and MSE depend on the measurement scale of the time-series
data. For example, forecasting profit in the range of millions of dollars would result in
very large values, even for accurate forecasting models. On the other hand, a variable
like market share, which is measured as a fraction, will always have small values of
MAD and MSE. Thus, the measures have no meaning except in comparison with other
models used to forecast the same data.
➢ MAPE is different in that the measurement scale factor is eliminated by dividing the
absolute error by the time-series data value. This makes the measure easier to
interpret. The selection of the best measure of forecasting accuracy is not a simple
matter; indeed, forecasting experts often disagree on which measure should be used.
However, MSE generally is the most popular.
Statistical Forecasting Models
➢ Forecasting methods can be classified as either statistical or judgmental.
➢ Statistical forecasting is based on the assumption that the future will be an
extrapolation of the past.
➢ Many different techniques exist; which technique should be used depends on
the variable being forecasted and the time horizon.
➢ Statistical methods can generally be categorized as time-series methods,
which extrapolate historical time series data, and regression methods, which
extrapolate historical time-series data but can also include other potentially
causal factors that influence the behavior of the time-series.
Statistical Forecasting Models
1. Simple Moving Average
2. Single Exponential Smoothing
3. Regression
Simple Moving Average
➢ This is based on the idea that averaging of random fluctuations in a time-series
to identify the underlying direction in which the time-series is changing.
➢ A moving average (MA) forecast is an average of the most recent “k”
observations in a time series. Thus, the forecast for the next period (t + 1), which
we denote as F (t+1), for a time series with t observations is

∑ (most recent “k” observations)

F
t+1 k
Simple Moving Average

∑ (most recent “k” observations)

F
t+1 k

( At + A t-1 + A t-2 + + A t-k + 1 )

k
Simple Moving Average
➢ MA methods work best for short planning horizons when there is no
major trend, seasonal, or business cycle patterns – that is, when demand
is relatively stable and consistent. As the value of k increases, the forecast
reacts slowly to recent changes in the time series because older data are
included in the computation.
➢ As the value of k decreases, the forecast reacts more quickly. If a
significant trend exists in the time-series data, moving average-based
forecasts will lag actual demand, resulting in a bias in the forecast.
Table 1.3
k=3
Illustration of Simple Moving Average
➢ Refer to table 1.3. The table shows the number of gallons of milk sold each
month at Gas-Mart, a local convenience store. To use moving averages to forecast
the milk sales, we must first select the number of data values to be included in the
moving average. As an example, let us compute forecasts based on a three-month
average (k=3). The moving average calculation for the first three months of the
milk-sales time series, and thus the forecast for month 4, is

172 + 217 + 190

F4
3
193.00
Illustration of Simple Moving Average
➢ The calculation of the second three month moving average F5 is

217 + 190 + 233

F5
3
213.33
➢ This provides a forecast for month 5. The error associated with this forecast is
179-213.33 = -34.33. The MSE for these forecasts is 1,457.33.
Moving Average
➢ The number of data values to be included in the moving average is often
based on managerial insight and judgment. Thus, it should not be
surprising that for a particular time series, different values of k lead to
different measures of forecast accuracy. One way to find the best number
is to use trial and error to identify the value of k that minimizes MSE for
the historical data.
THANK YOU!