Intr♥ducti♥n t♥

❤perati♥ns
Management

10
Forecasting

Linh Phuc – S1 23 24
Email: [email protected]
MSTeam: [email protected]
 a. Simple Moving
1. Naive Methods
average

1. Time-series 2. Techniques for b. Weighted

models Averaging moving average

3. Linear trend c. Exponential

A. Quantitative equation smoothing

1. Simple Linear
Regression
2. Associative
models
Forecasting 2. Multiple
techniques Regression
1. Executive
Opinions

2. Salesforce
Opinions
B. Qualitative
3. Consumer
Surveys

4. Delphi method

• Reading material: Chapter 3, Jay Heizer, Barry Render (2020). Operations Management, Sustainability and Supply
chain management, edition 13th. Pearson.

• Case study: P. 185 - 188, Chapter 4, Jay Heizer, Barry Render (2020). Operations Management, Sustainability and
Supply chain management, edition 13th. Pearson.
 a. Simple Moving
1. Naive Methods
average

2. Techniques for b. Weighted moving

1. Time-series models
Averaging average
Assumption:
The future is an
extension of the past
3. Linear trend c. Exponential
A. Quantitative  Historical data can be equation smoothing
forecasting models used to predict future
demand.
- Use mathematical 1. Simple Linear
techniques
- Based on historical data
Regression
- Use causal 2. Associative models
(explanatory) variables
- Less accurate as the 2. Multiple
Assumption:
forecast’s time horizon
One or more factors
Regression
increases
(independent variables)
are related to demand
They can be used to
predict future demand.

3. Forecast
Accuracy
 Trend variations Cyclical variations
- or  movements over many - Wavelike movements that
years are longer than a year
- Reason: factors such as - Reason: macroeconomic and
population growth, population
political factors.
shifts, cultural changes, and
income shifts. Time series
A time-ordered sequence of
observations taken at regular intervals.

Seasonal variations
- Peaks and valleys that repeat Random variations
over a consistent interval such
- Reason: force majeure
as hours, days, weeks, months,
years, or seasons.
 Explanation: the estimate for the next
period is equal to the actual demand for the
immediate past period.
 Formula:
𝐹𝑡 = 𝐴𝑡−1
where: 𝐹𝑡 = forecast for THIS period t
𝐴𝑡−1 = actual demand for the PREVIOUS period t-1
a) Simple Moving Average Forecast
 Explanation: uses historical data to calculate a moving average and
works well when the demand is fairly stable over time.
 Formula:
𝑡−1
1
𝐹𝑡 = × ෍ 𝐴𝑖
𝑛
𝑖=𝑡−𝑛

where: 𝐹𝑡 = forecast for THIS period t

𝐴𝑖 = actual demand for period i
n = No. of periods used to calculate moving average
t>n
b) Weighted Moving Average Forecast
 Explanation: An n-period weighted moving average forecast is the
weighted moving average of the n-period observations, using unequal
weights.
 Formula:
𝑡−1

𝐹𝑡 = ෍ 𝑤𝑖 𝐴𝑖
𝑖=𝑡−𝑛

where: 𝐹𝑡 = forecast for THIS period t

𝐴𝑖 = actual demand for period i
𝑤𝑖 = weight assigned to period i; 𝑤𝑖 >0; σ 𝑤𝑖 = 1
n = No. of periods used to calculate moving average
t>n
c) Exponential Smoothing Forecast
 Explanation: the forecast for the next period’s demand is the current
period’s forecast adjusted by a fraction of the difference between the
current period’s actual demand and forecast.
 Formula:
𝐹𝑡 = 𝐹𝑡−1 + 𝛼 𝐴𝑡−1 − 𝐹𝑡−1 = 𝛼𝐴𝑡−1 + (1 − 𝛼)𝐹𝑡−1

where: 𝐹𝑡 = forecast for THIS period t

𝐹𝑡−1 = forecast for the PREVIOUS period t-1
𝐴𝑡−1 = actual demand for the PREVIOUS period t-1
𝛼 = smoothing constant (0 ≤ 𝛼 ≤ 1)
 Explanation: using simple linear regression (trend line) to fit a line to
a series of data occurring over time.
 Formula: The trend line equation
𝑌෠ = 𝑏0 + 𝑏1 𝑡

where: 𝑌෠ = forecast or dependent variable

t = time variable, also independent variable values
𝑛 σ 𝑡𝑦 −σ 𝑡 σ 𝑦 σ 𝑦−𝑏1 σ 𝑡
𝑏1 = 2 ; 𝑏0 =
𝑛 σ 𝑡 2 −(σ 𝑡) 𝑛
y = dependent variable values
n = No. of observations
 Explanation: Just like the linear trend model, but the x variable is an
explanatory variable of demand, instead of time.
 Formula: The regression equation
𝑌෠ = 𝑏0 + 𝑏1 𝑥

where: 𝑌෠ = forecast or dependent variable

x = explanatory or independent variable values
𝑛 σ 𝑥𝑦 −σ 𝑥 σ 𝑦 σ 𝑦−𝑏1 σ 𝑥
𝑏1 = 2 ; 𝑏0 =
𝑛 σ 𝑥 2 −(σ 𝑥) 𝑛
y = dependent variable values
n = No. of observations
 Explanation: Just like the Simple Linear Regression Forecast, but
instead of 1, there are several explanatory variables.
 Formula: The multiple regression equation
𝑌෠ = 𝑏0 + 𝑏1 𝑥1 + 𝑏2 𝑥2 + ⋯ + 𝑏𝑘 𝑥𝑘

where: 𝑌෠ = forecast or dependent variable

𝑥𝑘 = kth explanatory or independent variable values
𝑏0 = constant
𝑏𝑘 = regression coefficient of the independent variable 𝑥𝑘
1) Forecast error is the difference between the actual quantity and
the forecast.
𝑒𝑡 = 𝐴𝑡 − 𝐹𝑡
where: 𝑒𝑡 = forecast error for period t
𝐴𝑡 = actual demand for period t
𝐹𝑡 = forecast for period t

2) Measures of forecasting accuracy (n = No. of errors)

σ𝑛𝑡=1 𝑒𝑡
𝑀𝑒𝑎𝑛 𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 𝑀𝐴𝐷 =
𝑛
𝑒
σ𝑛𝑡=1 𝑡
𝐴𝑡
𝑀𝑒𝑎𝑛 𝑎𝑏𝑠𝑜𝑙𝑢𝑡𝑒 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑎𝑔𝑒 𝑒𝑟𝑟𝑜𝑟 𝑀𝐴𝑃𝐸 =
𝑛
𝑛
σ𝑡=1 𝑒𝑡 2
𝑀𝑒𝑎𝑛 𝑠𝑞𝑢𝑎𝑟𝑒 𝑒𝑟𝑟𝑜𝑟 𝑀𝑆𝐸 =
𝑛
 1. Executive
Opinions

2. Salesforce
Opinions
B. Qualitative
3. Consumer
- Based on intuition or
judgmental evaluation Surveys
- Used to develop long-range
projections (current data is not
very reliable), and for new
product introductions (current
data is limited and/or
4. Delphi method
unavailable).
 Basically a meeting of senior
management executives to forecast the
market
 Apply for: long-range planning and new
product introductions; general demand
forecasting.
 Pros: They are knowledgeable and
experienced in their field, their forecast
will be valuable
 Cons: If one member’s views dominate
the discussion, then the value and
reliability of the outcome can be
diminished.
 Based on the knowledge of
sales team about the market
and estimates of customer
needs.
 Apply for: all kind of projects.
 Pros: The forecast tends to be
reliable because salespeople
are close to customers.
 Cons: Individual biases could
negatively impact the
effectiveness of this approach.
 • Design a forecasting questionnaire.
1

• Choose the target population.

2

• Carry out the survey through telephone, mail, Internet,

3 or personal interviews.

• Collect and analyze data.

4

• Make forecasts from the results.

5
 Accumulate,
Summarize and Send Announce final result
• Round 1 out results • Round n
• Experts • Round 2 • Experts • Final round
respond • Experts respond • Reach
Accumulate, respond Accumulated, consensus
Summarize and Send Summarize and Sent
out results out

 Apply for: high-risk technology forecasting; large, expensive projects;

or major new product introductions.
 Pros: Group members do not physically meet  avoid the scenario
where one or a few experts could dominate a discussion
 Cons: time-consuming and very expensive