UNIVERSITY OF TRANSPORT HOCHIMINH CITY

CHAPTER 2:
SUPPLY CHAIN ​OPERATIONS: PLANNING AND
SOURCING

Lecturer: ThS. Lê Quốc Lợi

Phone number: 0867 299 334

Mail: [email protected]

1
 CHAPTER 2

2.1. SUPPLY CHAIN ​OPERATIONS: PLANNING AND SOURCING

2.2. PLANNING
2.3. SOURCING

2
 CHAPTER 2 OBJECTIVES

After completing this chapter, students should achieve the following objectives:
✓ Presents an overview of planning and sourcing activities
✓ Using combine different methods to Demand Forecasting
✓ Analyzing Supply Chain Planning Activities
✓ Practicing Supply Chain Planning

3
 2.1 INTRODUCTION

PLANNING

• Demand Forecasting
• Product Pricing
• Inventory
Management

DISTRIBUTION SOURCING

• Order • Purchasing
Management • Payment
• Schedule Delivery
• Return Process

MANUFACTURING

• Product Design
• Production Process
• Facilities
Management

4
 2.2 PLANNING

- This is the process of planning a product from raw materials until it reaches the consumer.
- Includes:
+ Demand Forecasting
+ Supply Planning
+ Demand Planning
+ Manufacturing Planning
+ S&OP Planning

5
 2.2.1 Demand forecasting activities
- Demand forecasting is the basis for supply chain planning
- The accuracy of forecasts allows the supply chain to respond more quickly and efficiently
to customer needs
- Before selecting an appropriate forecasting method, businesses need to consider the
following issues:
• Past Demand
• Product Replenishment Timing
• Marketing or Advertising Plans
• Price Reduction Plans
• State of the Economy
• Competitor Activities

2. 3. 4.
1. Short-term forecasts are often Aggregate forecasts are often The farther from the customer,
Forecasts are always more accurate than long-term more accurate than individual the more distorted the
inaccurate forecasts. forecasts. information about their needs
becomes.

6
 Levels of demand forecasting

• Time: years
• Application: capacity forecasting, investment
Strategic
Forecast strategies, …

• Time: week, month, quarter

• Application: Sales planning, short-term budgeting,
Tactical
Forecast inventory planning, etc.

• Time: hours, days

• Application: decisions about production,
Operational transportation, replenishment of goods in the
Forecast warehouse

7
 Demand Forecasting Methods

Name Main content Advantages Disadvantages

- Effective with little historical data
Get opinions from experts, sales - Opinions may be subjective
Qualitative - Analyze diverse opinions
team, customers, etc. as a basis or limited to the participant's
- Suitable for new
for making forecasts. industries/products
level of understanding.

- Suitable with demand data

Based on historical customer does not vary greatly over - Not suitable for
Time series Analysis forecasting demand for
demand data to make forecasts. time
- Simple, easy to implement new products

Based on the correlation - Effective when the - Difficulty in finding factors

Causal correlation is clear and stable that correlate with
between demand and external
factors to make forecasts demand

Combining causal and time series

methods to simulate consumer - High precision - High implementation
Simulation
behavior under different
costs
hypothetical scenarios

8
 Demand Forecasting Methods

Name Main content Advantages Disadvantages

- Effective with little historical data
Get opinions from experts, sales - Opinions may be subjective
Qualitative - Analyze diverse opinions
team, customers, etc. as a basis or limited to the participant's
- Suitable for new
for making forecasts. industries/products
level of understanding.

- Suitable with demand data

Based on historical customer does not vary greatly over - Not suitable for
Time series Analysis forecasting demand for
demand data to make forecasts. time
- Simple, easy to implement new products

Based on the correlation - Effective when the - Difficulty in finding factors

Causal correlation is clear and stable that correlate with
between demand and external
factors to make forecasts demand

Combining causal and time series

methods to simulate consumer - High precision - High implementation
Simulation
behavior under different
costs
hypothetical scenarios

9
 Time series Analysis Methos

- Time series analysis is a widely used forecasting technique for forecasting demand for activities
with historical data.
- Common time series analysis methods include:
+ Naive method
+ Moving average method
+ Exponential smoothing method
+ Holt model method
+ Winter model method

10
 Components of a time series

(b) Trend (c) Seasonality

(a) Level

- Data fluctuates around a

value (average) - Data moves in a certain - Data increases and
- This is a fixed value if the data direction decreases in a repeating
does not contain the cycle over a period of
remaining factors time.
(d) Random factor (𝜺)
- Unable to determine displacement direction
- Predict size and variation to estimate error

11
 Steps to perform time series forecasting
Step 1: Data Analysis
From the recorded demand data, draw a demand-time correlation chart to find out the demand
characteristics based on the system components of the time series.

Step 2: Make the forecast

Make forecast for period t+1 with available data at period t using the following formula:
Ft+1 = Lt + lTt ∗ St

Step 3: Estimate the forecast deviation

Record the actual demand level in period t+1 (Dt+1) and compare it with the forecast level of
period t+1 (Ft+1) to determine the forecast deviation Et+1 , formula as below:
Et+1 = Ft+1 − Dt+1
Step 4: Calibration
Adjust the data and components of the time series in period t+1 after determining the forecast
deviation in period t+1. The adjusted data in period t+1 are used as the basis for making the
forecast for period t+2.

12
 Moving Average Model
- Used with the observed data has no trend or seasonal factors
- The forecast value is determined as the moving average of the observed data with a
predetermined number of calculation periods.

Formula:

Dt + Dt−1 + ⋯ + Dt−n+1
Ft+1 = Lt =
n

𝐹𝑡+1: forecast for period t+1

Lt : demand level in period t
𝐷𝑡−i : 𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 𝑑𝑒𝑚𝑎𝑛𝑑 𝑙𝑒𝑣𝑒𝑙 𝑖𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 t-i
n: number of observation periods

13
 Example
A milk tea shop reported weekly sales in August 2021 of 3Q oolong milk tea as follows:

Month Week Period (t) Demand (cup)

1 1 150
2 2 250
8
3 3 173
4 4 155

a, Use the moving average model with n = 4 to calculate the forecast demand for milk tea for the
first week of September.
b, If the sales in week 1 of September are 170 cups, what is the forecast deviation?

14
 Milk tea demand in August
300
Month Week Period (t) Demand (cup)
250
1 1 150
200
2 2 250
8 150
3 3 173
100

4 4 155 50

9 1 5 0
0 1 2 3 4 5

The demand levels for the 4 weeks in August are:

D4 +D3 +D2 +D1 155+173+250+150
L4 = 4
= 4
= 182 (cup)
a, Forecast of demand for milk tea in the first week of September is:
F5 = L4 = 182 (cup)
b, If the number of cups of milk tea sold in week 1 of September is 170 cups, the forecast
deviation is:
E5 = F5 − D5 = 182 − 170 = 12 (cup)
15
 Exercise 1
Weekly air conditioner sales in July and August 2022 of an electronics center are shown in the table
below:

Month Week Period (t) Demand (Unit)

7 4 1 98
1 2 143
2 3 115
8
3 4 125
4 5 100

a, Use the moving average model with n = 3 to calculate the forecast demand for air conditioners in the first
week of September.
b, If the number of air conditioners sold in the first week of September is 100 units, what is the forecast
deviation?

16
 Answers to exercise 1

Month Week Period (t) Demand (unit) Forcast

7 4 1 98
1 2 143
2 3 115
8
3 4 125
4 5 100
9 1 6 ?
The level of demand for goods in period 5 is:
D5 +D4 +D3 100+125+115
L5 = 3
= 3
= 113 (Unit)
a, The forecast demand for period 6 is:
F6 = L5 = 113 (Unit)
b, If the number of air conditioners sold in period 6 is 100 units, the forecast error in
period 6 will be:
E6 = F6 − D6 = 113 − 100 = 13 (Unit)
17
 Simple exponential smoothing model
- Used:
- The impact of data on customer demand increases over time → newer data has a higher value share
- Time series data has no trend or seasonality
- Leveling factor 𝛼 (0< 𝛼< 1) is used to represent the value of new data compared to old data

Formula:

Ft+1 = 𝛼. Dt + 1 − 𝛼 . Ft

Trong đó:
𝐹𝑡+1: Forecast period t+1
𝐷𝑡 : Actual level observed in the period t
F𝑡 : Forecast period t
𝛼: Smoothing coefficient satisfies the condition 0< 𝛼< 1

18
 Example
A milk tea shop reported weekly sales in August 2021 of 3Q oolong milk tea as follows:

Month Week Demand (cup)

1 150
2 250
8
3 173
4 155

Using a simple exponential smoothing model with 𝛼 = 0.1 to forecast the demand of milk tea
for the first week of September

19
 Milk tea demand in August
Đơn vị: ly
Month Week Period (t) Demand (cup) 300

1 1 150 250

200
2 2 250
8 150
3 3 173
100
4 4 155
50

9 1 5 0
0 1 2 3 4 5

Demand Forecast Demand

Month Week Period
(cup) (𝛼 = 0.1)
1 1 150 182
2 2 250 = 0.1 ∗ 150 + 0.9 ∗ 182 = 178.8
8
3 3 173 = 0.1 ∗ 250 + 0.9 ∗ 178.8 = 185.92
4 4 155 = 0.1 ∗ 173 + 0.9 ∗ 185.92 = 184.63
9 1 5 = 0.1 ∗ 155 + 0.9 ∗ 184.63 = 181.667

→ Forecast demand for milk tea in the first week of September is ……..cup
20
 Exercise 2
Weekly air conditioner sales in July and August 2022 of an electronics center are shown in the table
below:

Demand
Month Week Period (t)
Forecast (unit)
7 4 1 98
1 2 143
2 3 115
8
3 4 125
4 5 100
a, Use a simple exponential smoothing model with 𝛼 = 0.2 to calculate the forecast of air conditioner
demand in the first week of September.

21
 Answers to exercise 2

Month Week Period (t) Demand (Unit) Forecast (Unit)

7 4 1 98
1 2 143
2 3 115
8
3 4 125
4 5 100
9 1 6 ?

Demand
Month Week Period Demand Forecast (𝛼 = 0.2)
(Unit)
7 4 1 98 116.2
1 2 143 = 0.2 ∗ 98 + 0.8 ∗ 116 = 112.6
2 3 115 = 0.2 ∗ 143 + 0.8 ∗ 112.6 = 118.6
8
3 4 125 = 0.2 ∗ 115 + 0.8 ∗ 118.6 = 117.9
4 5 100 = 0.2 ∗ 125 + 0.8 ∗ 117.9 = 119.3
9 1 6 = 0.2 ∗ 100 + 0.8 ∗ 119 = 115.5
→ From the above calculation table, we have the forecasted demand for air conditioners for the first
week of September is 115 units.
22
 Holt Model
- Using the trend-adjusted exponential smoothing method
- Applicable when demand is trending but not seasonal
+ The trend and demand level estimates at period 0 are calculated using linear regression
(Dt=at+b), in which the demand level L0=b, trend factor T0 = a
+ Trend estimation in period t: Tt = 𝛽 ∗ Lt − Lt−1 + 1 − 𝛽 ∗ Tt−1
+ Estimated demand level in period t: Lt = 𝛼 ∗ Dt + 1 − 𝛼 ∗ (Lt−1 + Tt−1 )
+ The demand forecast for period t+1:

Ft+1 = Lt + Tt

𝐹𝑡+1 : Forecast of the period t+1

𝐷𝑡 : Actual demand level in the period t
L𝑡 : Stable demand level in the period t
T𝑡 : Trend level of the period t
𝛼: Smoothing coefficientsatisfies the condition 0< 𝛼< 1
𝛽: Smoothing coefficient tends to satisfy the condition 0<𝛽< 1

23
 Example

A milk tea shop reported weekly sales in August 2021 of 3Q oolong milk tea as follows:

Month Week Demand (cup)

1 150
2 170
8
3 200
4 250

Use Holt model with 𝛼 = 0.1, 𝛽=0.2 to forecast demand for bubble tea for the first week of
September

24
 Milk tea demand in August (unit: cup)
Month Week Period Demand x2 xy 300

(x) (y) 250

1 1 150 1 150 200

150
2 2 170 4 340
8 100
3 3 200 9 600
50
4 4 250 16 1000
0

σ 𝑥=10 σ y=770 σ x 2=30 σ xy=2090 0 1 2 3 4 5

4∗2090−10∗770
T0 = = 33
4∗30−102

770−33∗10
L0 = = 110
4

Using the regression equation showing the relationship between the demand variable and
the time variable, so: L0 = 110 và To = 33

25
 Period Demand Demand Level Demand
Month Week Trend (Tt)
(x) (y) (Lt) Forecast (Ft)
0 110 33
1 1 150 144 33 143
2 2 170 176 33 177
8
3 3 200 208 33 209
4 4 250 242 33 241
5 275

→ Forecasted demand for milk tea for the week 1: F1 = L0 + T0 = 110 + 33 = 143 (ly)
Estimated adjusted milk tea demand for the week 1: L1 = 0.1 ∗ 150 + 0.9 ∗ 143 = 144 (ly)
Estimated adjusted milk tea demand trend for the week 1: T1 = 0.2 ∗ 34 + 0.8 ∗ 33 = 33 (ly)
→ Forecasted demand for milk tea for the week 2: F2 = L1 + T1 = 144 + 33 = 177 (ly)
Similarly, the forecasted demand for milk tea in period 5 (week 1 of September) is 275 (cups)

26
 Exercise 3
Weekly air conditioner sales in July and August 2022 of an electronics center are shown in the table
below:

Month Week Period (t) Demand (unit)

7 4 1 98
1 2 143
2 3 140
8
3 4 150
4 5 160

a, Use the Holt model with 𝛼 = 0.1, 𝛽 = 0.2 to calculate the forecast of air conditioner demand in the first
week of September.

27
 Answers to exercise 3
Month Week Period (t) Demand (unit) x2 xy

7 4 1 98 1 98
1 2 143 4 286
2 3 140 9 420
8
3 4 150 16 600
4 5 160 25 800
σ 𝑥=15 σ y=691 σ x 2 =55 σ xy=2204

5∗2204−15∗691
T0 = = 13.1
5∗55−152

691−13.1∗15
L0 = = 98.9
5

Use regression equation to show the relationship between demand variable and time
variable, so: L0 = 98.9 và To = 13.1

28
 Demand
Month Week Period (t) Demand level (Lt) Trend (Tt) Forecast (Ft)
(unit)
0 98.9 13.1
7 4 1 98 =0.1×98+(1-0.1) ×112= 110.6 =0.2×(110.6-98.9)+(1-0.2) ×13.1=12.8 =98.8+13.1 = 112
1 2 143 =0.1×143+(1-0.1) ×123.4=125.4 =0.2×(125.4-110.6)+(1-0.2) ×12.8=13.2 =110.6+12.8=123.4
2 3 140 =0.1×140+(1-0.1) ×138.6=138.7 =0.2×(138.7-125.4)+(1-0.2) ×13.2=13.2 =125.4+13.2=138.6
8
3 4 150 =0.1×150+(1-0.1) ×152=158.1 =0.2×(158.1-138.7)+(1-0.2) × 13.2=13.2 =138.7+13.2=152
4 5 160 =0.1×160+(1-0.1) ×165=164.5 =0.2×(164.5-158.1)+(1-0.2) ×13.2=13.1 =158.1+13.2=165
6 =164.5+13.1=177.6

→ Forecasted demand for air conditioners for the week 1: F1= L0 + T0 = 98.9 + 13.1 = 112 (unit)
Estimated demand for air conditioners adjusted for the week 1: L1 = 0.1×98 + 0.9×112 = 110.6 (unit)

Estimated adjusted milk tea demand trend for the week 1: T1 = 0.2×34 + 0.8×13.1 = 12.8 (unit)
→ Forecasted demand for air conditioners for the week 2: F2 = L1 + T1 = 110.6 + 12.8 = 123.4 (unit)
Similarly calculated, the forecast demand for air conditioners for week 5 is 178 units.

29
 Winter Model
- Used with demand is both trending and seasonal

+ Estimate the stability of demand in the period t+1: Lt+1 = 𝛼 ∗ (Dt+1 ΤSt+1 + 1 − 𝛼 ∗ (Lt + Tt )
+ Estimate the trend of demand in the period t+1: Tt+1 = 𝛽 ∗ Lt+1 − Lt + 1 − 𝛽 ∗ Tt
+ Estimated seasonality of demand in the period t+1: St+p+1 = 𝛾 ∗ Dt+1 /Lt+1 + 1 − 𝛾 ∗ St+1
+ Demand forecast for period t+1 is:

Ft+1 = (Lt +Tt ) ∗ St+1

𝐹𝑡+1 : Level forecast in period t+1

𝐷𝑡 : Actual demand level in the period t
L𝑡 : Stable demand level in the period t
T𝑡 : Trend level of the period t
𝛼: Smoothing coefficient (0< 𝛼< 1)
𝛽: Trend smoothing coefficient (0<𝛽< 1)
𝛾: Seasonal smoothing coefficient (0 < 𝛾<1)

30
 Choosing the right forecasting model
Method 1: Based on the characteristics of the needs

Forecasting method Applicable cases

Moving average Demand is not seasonal or trend-driven.

Simple exponential smoothing Demand is not seasonal or trend-driven.

Holt Model Demand is trending but not seasonal.

Winter Model Demand is both trend and seasonality.

Method 2: Based on comparing the deviation of forecasting methods

31
 Measuring forecast deviation (Et = Ft − Dt )
Why do we need measure forecast deviation?
+ Determine the accuracy of the forecasting method and adjust the method when necessary
+ Basis for contingency plans

Measurement index Formula Application

Mean square deviation σnt=1 Et2 Measurement when deviation

MSEn =
- MSE n are symmetrically distributed
σnt=1 A t
MADn = Measuring deviation when
Mean absolute deviation - MAD n error has asymmetric
distribution
(𝐴𝑡 = 𝐸𝑡 )
Et Measure forecast deviation
σnt=1 100
Dt effectively when forecasts are
Average absolute percent error - MAPE MAPEn =
n seasonal and demand varies
significantly across periods

32
 Exercise 4
Weekly air conditioner sales in July and August 2022 of an electronics center are shown in the table below:

Month Week Period Demand Forecast

7 4 1 98 112
1 2 143 123
2 3 140 139
8
3 4 150 152
4 5 160 165

Calculate MSE, MAD and MAPE in above case.

33
 Answers to exercise 4

Devi
Period Demand Forecast
ation Et2 MSEt |Et| MADt |%Et| MAPEt
(t) (Dt) (Ft)
(Et)
1 98 112 14 196 196 14 14 14% 14%
2 143 123 -20 400 298 20 17 14% 14%
3 140 139 -1 1 199 1 12 1% 10%
4 150 152 2 4 150 2 9 1% 8%
5 160 165 5 25 125 5 8 3% 7%

Based on above table, so:

• MSE5 = 125
• MAD5 = 8
• MAPE5 = 7%

34
 Exercise 5
Weekly demand at Hot Pizza is shown in the Questions:
table below:
a. Forecast demand for the next 4 weeks using a moving
Month Demand ($) average model with n=4 and a simple exponential
smoothing model with smoothing factor 𝛼 = 0.1
1 110
2 118 b. b. Calculate MAD, MAPE, MSE in any single case.
3 119 c. In your opinion, which of the two forecasting models above
4 134 gives more accurate results in this case? Why?
5 92
6 115
7 90
8 106
9 118
10 106
11 95
12 93

35
 Exercise 6
Monthly demand for flat panel TVs at A&D Questions:
Electronics is shown in the table below:
a. Forecast the demand for the next 2 weeks using the simple
exponential smoothing model with coefficient 𝛼 = 0.3 and
Month Demand (Unit) the Holt model with coefficient 𝛼 = 0.05 and coefficient 𝛽 =
1 1,000 0.1
2 1,113
b. Note:
3 1,271
• With the simple exponential smoothing model, the
4 1,445
demand level at time t0 (L0): 1,659 unit (average demand
5 1,558 for 12 months)
6 1,648
• With the Holt model, the demand level at time t=0 (L0):
7 1,724
948 unit and the trend factor of demand at the time t=0
8 1,850 (T0): 109 unit.
9 1,864
b. Calculate MAD, MAPE, MSE in any single case.
10 2,076
11 2,167
c. In your opinion, which of the two forecasting models above
gives more accurate results in this case? Why?
12 2,191

36