Forecasting

Learning objectives
Forecasting is essential
Quantitative forecasting models
Forecasting error
Qualitative forecasting
Collaborative forecasting
Forecasting in OSCM
• Vital for every business decision
• For accounting, finance forecasts are the basis for budgetary planning
and cost control
• For marketing-new product decision, compensating sales personnel,
marketing communications
• Production & operations- supplier selection, process selection,
capacity planning, lay-out, purchasing, production scheduling,
inventory
• Approach to forecast is determined by the purpose of forecast
Approaches to forecast
• Strategic forecast
How to meet the demand
For decisions related to process design, capacity planning, sourcing, location
and distribution planning
Medium to long-term
Estimate aggregate demand
• Tactical forecast
Ensure product/service availability given the lead-time expectations of the
customers
Weekly or monthly demand forecast
Quantitative forecasting models
• Time Series Analysis
• Data relating to past demand can be used to predict future demand.
• Past data may include several components, such as trend, seasonal, or cyclical
influences
• Causal Relationship Analysis
• Certain economic, social, demographic or other factors influence demand
• Regression analysis is done
• Simulation
• Based on large data set estimate the distribution of the variables that
influence demand
• Estimate future demand using simulated data sets
Components of Demand
Six components:
1. average demand for the period,
2. a trend,
3. seasonal element,
4. cyclical elements (political, economical
factors, war, sociological issues may exert
cyclicality in demand),
5. random variation (chance events), and
6. Auto-correlation (persistence of
occurrence)
Common trends
Selecting appropriate forecasting method
Fore Cast Method Amount of Data Data Pattern Horizon
Simple moving average 6 to 12 data points (weekly/monthly) Stationary (no trend) Short
Weighted moving average 5 to 10 data points Stationary Short
Simple exponential smoothing 5 to 10 data points Stationary Short
Exponential smoothing with trend 5 to 10 data points Stationary and trend Short
Linear regression
10 to 20 data points Stationary, trend and Short/Med
Trend and seasonal model seasonality
2 to 3 observations per season Stationary, trend and Short/med
seasonality
Short term (less than three month) for tactical decision e.g., setting
safety stock level, estimating peak load, respond to random Which forecasting model a firm
variations etc. should choose depends on:
Medium term ( 3 month to 2 years) for planning a strategy for Time horizon to forecast
meeting demand over the next six months to a year and a half; useful Data availability
for capturing seasonal effects. Accuracy required
Long term (more than 2 years) detect general trends and are Size of forecasting budget
especially useful in identifying major turning points. Availability of qualified personnel
Simple and weighted moving average
Choice of time period depends on the
purpose of the forecast.
Monthly data for budgeting
Weekly data for production
scheduling or inventory planning

Shorter periods in the average allow the

ups and downs to persist in the
forecasted data; whereas longer periods
in the average smooth out the ups and
downs (random variations) in data set.

Choosing weights in case of weighted

moving average- general rule is to assign
more weight to the most recent data
 Exponential smoothing
• Moving average’s drawback is over
reliance on historical data
• But in most contexts, the importance of
data diminishes as the past becomes more
distant
• Only three pieces of data are needed: the
most recent forecast, the actual demand
that occurred for that forecast period, and
a smoothing constant alpha (α).
• The value of the smoothing constant is
determined both by the nature of the
product and by the manager’s sense of
what constitutes a good response rate. It is
the reaction rate to forecast error of the
immediate past period. High value is
suggested in case of volatile demand.
Exponential smoothing with trend
• Exponential forecasts always lag behind the trends in data set

Choosing the Appropriate Value for Alpha and Delta

Typically, fairly small values are used for alpha and delta in the range of 0.1 to 0.3.
The values depend on how much random variation there is in demand and
How steady the trend factor is.
Regression
• Linear regression is useful for
long-term forecasting of major
occurrences and aggregate
planning.
• The data should be plotted
first to see if they appear
linear or if at least parts of the
data are linear.
• Yt= a + bt
Linear Regression
Decomposition of a time series
Computing trend and seasonal index
Forecasting causal relationship
Using independent variables other than time to predict future demand
Forecast error
• Sources of error
• Random (unexplained)
• Biases (in the data set, trend and/or seasonality identification, unexpected
change in trend/seasonality)
• Measurement error
• Error in identifying the causal variables
• Measurement of error
• MAD
• MAPE
• Tracking Signal
Measurements of forecast errors
Qualitative Techniques in forecasting
Panel Consensus
Panel forecasts are developed through open meetings with a free exchange of ideas from all levels of
management and individuals. The difficulty with this open style is that lower-level employees are
intimidated by higher levels of management.
Historical Analogy
In trying to forecast demand for a new product, an ideal situation would be where an existing product or
generic product could be used as a model.
The Delphi Method
As we mentioned under panel consensus, a statement or opinion of a higher-level person will likely be
weighted more than that of a lower-level person. To prevent this problem, the Delphi method conceals
the identity of the individuals participating in the study. The step-by-step procedure for the Delphi
method is:
Choose the experts to participate. There should be a variety of knowledgeable people in different
areas.
Through a questionnaire (or e-mail), obtain forecasts from all participants.
Summarize the results, and redistribute them to the participants along with appropriate new
questions.
Summarize again, refining forecasts and conditions, and again develop new questions.
Repeat step 4 if necessary. Distribute the final results to all participants.
Collaborative forecasting
• Collaborative Planning, Forecasting, and Replenishment (CPFR) is a Web-based tool
used to coordinate demand forecasting, production and purchase planning, and
inventory replenishment between supply chain trading partners.
• CPFR is being used as a means of integrating all members of an n-tier supply chain,
including manufacturers, distributors, and retailers.
• The ideal point of collaboration utilizing CPFR is the retail-level demand forecast,
which is successively used to synchronize forecasts, production, and replenishment
plans upstream through the supply chain.
Exercises
Sunrise Baking Company markets doughnuts through a chain of food stores. It has been experiencing
overproduction and underproduction because of forecasting errors. The following data are its demand in
dozens of doughnuts for the past four weeks. Doughnuts are made for the following day; for example,
Sunday’s doughnut production is for Monday’s sales, Monday’s production is for Tuesday’s sales, and so
forth. The bakery is closed Saturday, so Friday’s production must satisfy demand for both Saturday and
Sunday. Make a forecast for this week on the following basis:
• Daily, using a simple four-week moving average.
• Daily, using a weighted moving average with weights of 0.40, 0.30, 0.20, and 0.10 (most recent to oldest
week).
• Sunrise is also planning its purchases of ingredients for bread production. If bread demand had been
forecast for last week at 22,000 loaves and only 21,000 loaves were actually demanded, what would
Sunrise’s forecast be for this week using exponential smoothing with α = 0.10?
• Suppose, with the forecast made in part (c), this week’s demand actually turns out to be 22,500. What
would the new forecast be for the next week?
Exercises
Given the following information, make a forecast for May using exponential
smoothing with trend. Make a second forecast using linear regression.

For exponential smoothing with trend, assume that the previous forecast (for
April) including trend (FIT) was 800 units, and the previous trend component
(T) was 50 units. Also, alpha(α) = 0.3 and delta(δ) = 0.1.
For linear regression, use the January through April demand data to fit the
regression line. Use the Excel regression functions SLOPE and INTERCEPT to
calculate these values.
Exercises
Here are the quarterly data for the past two years. From these data,
prepare a forecast for the upcoming year using a linear regression with
seasonal indexes.
Year- Quarter Sales
1-Q1 300
1-Q2 540
1-Q3 885
1-Q4 580
2-Q1 416
2-Q2 760
2-Q3 1191
2-Q4 760