Professional Documents
Culture Documents
SCM 08 - Periodic Inventory Managament
SCM 08 - Periodic Inventory Managament
SCM 08 - Periodic Inventory Managament
2
Inventory Management Policies:
Introduction
(R,Q)-policy (T,Q)-policy
Fix order quantity
Q e.g.: Fuel: fill up 30 litres as soon as the e.g.: only in case of constant demand (drugs)
inventory has reached a critical value. e.g.: Fuel: 30 litres every 4 days.
(R,S)-policy
Variable order (T,S)-policy
quantity e.g.: Fuel: fill up to the maximum load as
e.g.: Fuel: fill up to the maximum load every 4
S soon as the inventory has reached a critical
days.
value.
+ low expenditures
Advantages + low risk of shortages + combination of ordering with other items
possible (lowers fix costs per item)
3
Periodic Multi-Period Inventory Management
The (T,S)-Policy or the „Order-up-to Model“
CT-14
Single Period vs. Multi Period Model
Lead time = 1:
5
Multi-Period Inventory Management
Model Assumptions
1. The firm is active during multiple periods.
2. Orders are placed at the beginning of each period, where T is called the review
interval, as the time (e.g. in weeks or days) between two successive orders.
3. The Distribution 𝐹(𝐷) of demand 𝐷 is known. Its mean value equals 𝜇 and its
standard deviation is 𝜎. (We mainly assume Poisson or Normal distributions, but any
distribution can be used.)
4. Variable costs per ordered item = 𝑐
5. Fix ordering costs are negligible 𝐾 = 0
6. Surplus inventory can be used to satisfy demand of future periods.
7. Inventory holding costs equal ℎ per item per period. Inventory holding costs have
to be considered for every item which is in stock at the end of the period.
8. Backorders are allowed. Stock-out costs (penalties) 𝑝 have to be paid per missing
item per period.
9. Orders are delivered using a lead time 𝐿𝑇 (e.g. weeks or days).
10. There are no restrictions concerning the order quantity.
11. Decision: Find the target inventory 𝑆, minimize the expected costs.
6
Target inventory policy 𝑺
• Under our model assumptions, the target inventory policy is optimal. This policy
assumes, that we order the difference of the current inventory level 𝐼𝑡 (plus
outstanding orders) and the target inventory 𝑆 at the beginning of each period 𝑡.
𝐼𝑡
Given ℎ = 1 and 𝑝 = 3, the total 50
costs are: 40
0 + 10 + 0 + 20 ∙ 1 = 30 20
• Stockout costs: 10
10 ∙ 3 = 30
• Total costs: 30 + 30 = 60 0
1 2 3 4
-10
8
Target inventory policy 𝑺
Simple Example, 𝑆 = 50, 𝐿𝑇 = 0
𝐼𝑡
Given ℎ = 1 and 𝑝 = 3, the total 50
costs are: 40
0 + 20 + 0 + 30 ∙ 1 = 50 20
• No Stockout costs 10
• Total costs: 50 + 0 = 50
0
1 2 3 4
-10
9
Order up-to model implementation
• What are the costs of having ordered one unit too much in the
given period? holding costs ℎ
• What are the costs of having ordered one unit too little in the
given period? penalty costs 𝑝
11
Target inventory policy 𝑺
Determining the cost and profit functions, 𝑳𝑻 = 𝟎
𝑝 𝐶𝑢
Critical Ratio 𝐶𝑅 = 𝐶𝑅 =
𝑝+ℎ 𝐶𝑢 + 𝐶𝑜
Optimal Order-
up-to level and 𝑆∗ = 𝜇 + 𝑧 ∙ 𝜎 𝑄∗ = 𝜇 + 𝑧 ∙ 𝜎
Optimal Order with 𝑧 = 𝐹𝑁−10,1 (𝐶𝑅) with 𝑧 = 𝐹𝑁−10,1 (𝐶𝑅)
Quantity
12
Target inventory policy 𝑺
Determining the cost and profit functions, 𝑳𝑻 = 𝟎
13
Target inventory policy 𝑺
How to deal with 𝑳𝑻 > 𝟎
• Notation
𝑡 Time period
𝐿𝑇 Lead time in periods between ordering and receiving units.
𝐼𝑡 Physical inventory level at the beginning of period 𝑡
𝑂𝑡 Outstanding (on-order) inventory level at beginning of period 𝑡
Disposable inventory level at the beginning of period 𝑡 (before
𝐼𝑃𝑡
ordering in period 𝑡) = 𝐼𝑡 + 𝑂𝑡
𝑆 Target inventory level
𝑋𝑡 Order quantity at the beginning of period 𝑡 ⇒ 𝑋𝑡 = 𝑆 − 𝐼𝑃𝑡
𝐷𝑡 Demand in period 𝑡.
16
Target inventory policy 𝑺
How to deal with 𝑳𝑻 > 𝟎, simple example, 𝑺 = 𝟏𝟓, 𝑳𝑻 = 𝟐
• Total costs: 45 -4
-5
17
Target inventory policy 𝑺 with 𝑳𝑻 > 𝟎
Determining the cost and profit functions
18
Target inventory policy 𝑺
Determining the cost and profit functions
𝑝 𝐶𝑢
Critical Ratio 𝐶𝑅 = 𝐶𝑅 =
𝑝+ℎ 𝐶𝑢 + 𝐶𝑜
Optimal Order-
up-to level and 𝑆 ∗ = 𝜇𝐿𝑇+𝑇 + 𝑧 ∙ 𝜎𝐿𝑇+𝑇 𝑄∗ = 𝜇 + 𝑧 ∙ 𝜎
Optimal Order with 𝑧 = 𝐹𝑁−10,1 (𝐶𝑅) with 𝑧 = 𝐹𝑁−10,1 (𝐶𝑅)
Quantity
19
Target inventory policy S,
Performance Measures
For any target inventory policy („order-up-to level“) S, we can evaluate the
following performance measures:
• Expected backorders (EB): Expected number of back orders at the end of
each period.
• Safety inventory (SI): Inventory which is carried to absorb higher than
expected demand.
• Expected „leftover“ inventory (ELI): Expected inventory level at the end of
the review period.
• Expected ordering quantity (EQ)
• Expected average inventory (EAvI): Expected average inventory during the
review period.
• Expected costs (EC): Costs during one review period.
• Expected fill rate (EFR, 𝜷): The fraction of demand that is satisfied directly
from stock.
• In-stock probability or cycle service level (CSL, 𝜶): Probability all demand
during the review period is directly delivered (from stock).
20
Target inventory policy S,
Performance Measures
For any target inventory policy („order-up-to level“) 𝑆, we can
evaluate the following performance measures:
• Expected backorders (EB): 𝐸𝐵 = 𝐿 𝑧 ∙ 𝜎𝐿𝑇+𝑇
• Safety inventory (SI): 𝑆𝐼 = 𝑧 ∙ 𝜎𝐿𝑇+𝑇
• Expected „leftover“ inventory (ELI): 𝐸𝐿𝐼 = 𝑆 − 𝜇𝐿𝑇+𝑇 + 𝐸𝐵 =
𝑆𝐼 + 𝐸𝐵
• Expected ordering quantity (EQ): 𝐸𝑄 = 𝜇 𝑇
• Expected average inventory (EAvI): 𝐸𝐴𝑣𝐼 = 𝐸𝑄/2 + 𝑆𝐼
• Expected costs (EC1): 𝐸𝐶1 = ℎ ∙ 𝐸𝐴𝑣𝐼 + 𝑝 ∙ 𝐸𝐵
• Expected costs (EC2, CT): 𝐸𝐶2 = ℎ ∙ 𝐸𝐿𝐼 + 𝑝 ∙ 𝐸𝐵
𝐸𝐵
• Expected fill rate (𝛽): 𝛽 = 1 −
𝜇𝑇
• In-stock probability or cycle service level (𝛼): 𝛼 = 𝐹(𝑧)
21
Target inventory policy S,
Example for 𝑳𝑻 = 𝟑, 𝑻 = 𝟏
• 𝜇 = 200, 𝜎 = 50, 𝑐 = 4.50, 𝑟 = 7.00, 𝑝 = 1.00, ℎ = 0.05, 𝑇 = 1
• 𝜇𝐿𝑇+𝑇 = 800, 𝜎𝐿𝑇+𝑇 = 3 + 1 ∙ 50 = 100
𝑝 1.00
• 𝐶𝑅 = 𝑝+ℎ = 1.00+0.05 = 0.9524 ⇒ 𝑧 = 1.67 !
• 𝑆 ∗ = 𝜇𝐿𝑇+𝑇 + 𝑧 ∙ 𝜎𝐿𝑇+𝑇 = 800 + 1.67 ∙ 100 = 967
• Expected backorders (EB): 𝐸𝐵 = 𝐿 𝑧 ∙ 𝜎𝐿𝑇+𝑇 = 0.0197 ∙ 100 = 1.97 ≅ 2
• Safety inventory (SI): 𝑆𝐼 = 𝑧 ∙ 𝜎𝐿𝑇+𝑇 = 1.67 ∙ 100 = 167
• Expected „leftover“ inventory (ELI): Should be plus+
22
Target in-stock probability 𝜶
𝜶-Servicelevel
• As with the newsvendor model, we could aim to reach a minimum in-
stock probability instead of maximizing cost.
• Remember, that 𝐼𝑡 = 𝑆 − 𝐷𝐿𝑇+𝑇
• Then, we choose lowest 𝑆 ∗ , so that
𝑃 𝐼𝑡 ≥ 0 = 𝑃 𝑆 − 𝐷𝐿𝑇+𝑇 ≥ 0 = 𝑃 𝐷𝐿𝑇+𝑇 ≤ 𝑆 = 𝐹𝐿𝑇+𝑇 𝑆 = 𝛼
• Choose lowest 𝑺∗ and lowest 𝒛 so that the desired level in-stock
probability is reached.
23
Target fill-rate 𝜷
𝜷-Servicelevel
𝜎𝐿𝑇+𝑇 ∙𝐿(𝑧)
• Recall the fill-rate equals = 1 − .
𝜇𝑇
𝜎𝐿𝑇+𝑇 ∙𝐿(𝑧)
• Then, we choose lowest 𝑆 ∗ , so that 1 − ≥ 𝛽.
𝜇𝑇
• If demand is normally distributed, 𝐿(𝑧) can be found within the
corresponding Loss Function Table:
𝜎𝐿𝑇+𝑇 ∙ 𝐿(𝑧) 𝜇𝑇
≤ 1 − 𝛽 ⇒ 𝐿(𝑧) ≤ (1 − 𝛽) ∙
𝜇𝑇 𝜎𝐿𝑇+𝑇
• Choose the lowest 𝒛 and correspondingly lowest 𝑺∗ , so that the above
inequality still holds.
• E.g. assume (𝐿𝑇 = 1, 𝑇 = 1) 𝐷𝑡 ~𝑁 200, 50 and
𝐷𝐿𝑇+𝑇 ~𝑁 400, 70.71 and 𝛽 = 0.98:
200
𝐿 𝑧 ≤ 1 − 0.98 ∙ ⇒ 𝐿 𝑧 ≤ 0.05657 ⇒ 𝑧 = 1.20
70.71
𝑆 ∗ = 𝜇𝐿𝑇+𝑇 + 𝑧 ∙ 𝜎𝐿𝑇+𝑇 = 400 + 1.20 ∙ 70.71 = 484.852 ≅ 485
24
Period length 𝑻 and lead time 𝑳𝑻
• 𝑇 = 2, 𝐿𝑇 = 8 weeks
LT
P1 P2 P3 P4
• 𝑇 = 4, 𝐿𝑇 = 8 weeks
LT
P1 P2
• 𝑇 = 8, 𝐿𝑇 = 8 weeks LT
P1
25
Period length 𝑻 and lead time 𝑳𝑻
• Assume 𝐿𝑇 = 8 weeks. Then the period length 𝑇 could be either {1, 2, 4, 8} weeks. Assume fixed
ordering costs 𝐾 = 0. Does this make a difference in our setting? Yes!
◦ Again, demand for one week ~𝑁(200,50), ℎ = 0.05, 𝑝 = 1.00
T [weeks] 1.00 2.00 4.00 8.00
LT [weeks] 8.00 8.00 8.00 8.00
mu_week 200.00 200.00 200.00 200.00
sigma_week 50.00 50.00 50.00 50.00
mu_(T+LT) 1 800.00 2 000.00 2 400.00 3 200.00
sigma_(T+LT) 150.00 158.11 173.21 200.00
CSL 0.9990 0.9990 0.9990 0.9990
z 3.0902 3.0902 3.0902 3.0902
L(z) 0.0003 0.0003 0.0003 0.0003
S 2 263.53 2 488.61 2 935.24 3 818.05
EQ 200.00 400.00 800.00 1 600.00
SI 463.53 488.61 535.24 618.05
EAvI 563.53 688.61 935.24 1 418.05
EB 0.04 0.04 0.05 0.06
ELI 463.58 488.65 535.29 618.10
EC1 28.22 68.90 187.10 567.27
EC2 23.22 48.91 107.11 247.30
EC1 - 52 weeks 1 467.35 1 791.52 2 432.26 3 687.28
EC2 - 52 weeks 1 207.46 1 271.63 1 392.38 1 607.42
26
Period length 𝑻 and lead time 𝑳𝑻
• Assume 𝐿𝑇 = 8 weeks. Then the period length 𝑇 could be either {1, 2, 4, 8} weeks. Assume fixed
ordering costs 𝐾 = 0. Does this make a difference in our setting? Yes!
27