INVENTORY MODELS

Probabilistic

Ms. Giovana Valverde Ayala

Probabilistic Inventory Models

They are inventory models in which demand is
uncertain or random in a given period: period model
unique, model for various periods, and review policies
continues. When the assumption of constant demand
it cannot be sustained, it is possible to use some models in
those that the demand is a random variable with distribution
of known probability.
In these cases, the decision-maker is indifferent to the risk.
and assess your alternatives based on the average of the
results that they provide in an extensive number of
repetitions of the same situation.

Ms. Giovana Valverde Ayala 2

1
 Probabilistic Inventory Models
Given the conditions, a model is sought to be developed that
adjust to the situation. Only some models will be analyzed
with the intention that they constitute a practice for the
construction of what is needed on each occasion.
The models have been designed to address problems of
stock of goods. The probability distributions
They can assume continuous or discrete variations.
An important concern of the management is the
maintenance of an adequate level of service when facing
an uncertain demand. The level of service is the complement
of the probability of a missing item.

Ms. Giovana Valverde Ayala 3

Probabilistic Inventory Models

from a single period
Single-period inventory models occur when a
the article is ordered once, solely to satisfy the
demand for a specific period. For example, an item from
fashion becomes obsolete after a certain period and then
it cannot be requested again.
Los modelos de un solo periodo se investigaran en condiciones
different, mainly including instant demands and
uniforms with a fixed cost and without it. It is assumed that the
replenishment occurs instantly. The inventory level
optimum is deduced based on cost minimization
expected inventory, which includes order costs (cost
fixed plus cost of purchase or production) of maintenance and
scarcity. As demand is probabilistic, the cost per unit of
purchase (production) will be an effective factor in the cost function.

Mg. Giovana Valverde Ayala 4

2
Probabilistic Inventory Models

What quantity of stock should

maintain the company to protect oneself
reasonably against uncertainty?
The probabilistic models to be presented are
they consider a period or multiple periods.
It is not advisable to allow the level of
stocks reach the level of zero.

Mg. Giovana Valverde Ayala 5

There are uncertainty factors regarding:

Variable demand.
Waiting time.
Demand during the waiting time.

VariableDemand
The demand future he uncertain.
Information must be gathered regarding situations
past data to estimate future demand.
The mean and the standard deviation are calculated for
obtain demand indicators, using the points
average of the intervals.

Ms. Giovana Valverde Ayala 6

3
 Waiting time
The waiting time to receive orders, just like
demand is often uncertain and this behavior
is distributed in some way.

Demand during the waiting time

The two sources of variation in demand during the
waiting time is:
The duration of the waiting time, by period.
The demand in the waiting time, by period.
They interact to determine the demand during the
wait time

Ms. Giovana Valverde Ayala 7

Probabilistic Model of a period without

Initial stock

Planning is carried out for a period.

The demand of a random variable with a
probability distribution.
There is no initial inventory.
The decision to be made is the value of Y, the number of
units that need to be bought or produced
start of the period for the inventory.

Ms. Giovana Valverde Ayala 8

4
 Objetivo
Minimize the expected total cost, where the components
of the cost are:

C=Unit cost of buying or producing each unit.

h = Cost of holding per unit of inventory
remains at the end of the period (includes the cost of
storage minus the recovery value.
P = Cost of penalty for missing unit during
the period.
A= Fixed Cost

Mg. Giovana Valverde Ayala 9

D=Probabilistic demand during the period.

f(D) = fdp of the demand during the period.

Y= Order quantity.

The model determines the optimal value of Y that

minimize the sum of expected purchase costs
(or production), management and shortage.

Ms. Giovana Valverde Ayala 10

5
 ANALYSIS OF THE MODEL
The decision on the value of Y (inventory quantity)
to acquire), depends heavily on the distribution of
probability of the demand D.
It may be desirable to exceed the expected demand.
but perhaps not reaching the maximum possible demand.
It is necessary to have a balance or a trade-off between:
1. The risk of a shortage that involves costs for
missing.
2. The risk of having a surplus and incurring in the
wasted costs of ordering and storing
more than the necessary units.

Ms. Giovana Valverde Ayala 11

GRAPHIC

D<Y D>Y

Y D

Y-D Y
D

D-Y

h (Y-D), if D ≤Y
0 , if Y=D p (D-Y), if D ≥ Y
0 , if D=Y

Mg. Giovana Valverde Ayala 12

6
 This is achieved by minimizing the expected value (in the
statistical sense) of the sums of these costs.
The quantity sold is given by:

D is less than y
Min {D,Y} = And if D>y

Thus, the Demand isDy and it is storedY, the cost

incurred is given by:

C(D, y) = cy + pmax{0, D - y} + hmax{0, y - D} + A

Mg. Giovana Valverde Ayala 13

As demand is a random variable [with the

probability distribution f(D) dDthis cost is also
a random variable.

The expected cost is given by C(y), where:

For the continuous case:

C(y) = E{C(y)}= [cy + hmáx{0, y - D} + pmáx{0,D-y} + A] f(D)dD

Ms. Giovana Valverde Ayala 14

7
E{C(y)}= Costo de ordenar + E{costo de mantener} + E{costo de escasez}
y
∞
E{C(y)} = Cy + h [(y -D)]f(D)dD+ p y[(D- y)]f(D)dD + A
d=0

E{C(y)} = Cy + h [(y - D)]f(D)dD - p [(y -D)]f(D)dD+ A

= cy + A + L(y)

Where L(y) is known as the expected shortage cost

and storage.

Ms. Giovana Valverde Ayala 15

It is necessary to obtain the value of y that minimizes C(y).

To obtain the value that minimizes the expected cost

(C(y)) we must take into account the following:

Let D be a random variable that has a function of

densidad :
f(D), if D ≥ 0
0, in another way
y
It is denoted P {D ≤ y} = f(D) dD
0

Ms. Giovana Valverde Ayala 16

8
2. Derive the expected cost C(y) with respect to Y
and set to zero: y
d (E{C(y)}) = C + h f(D) dD - p f(D) dD = 0
d=0 d=Y
dy

This expression implies that:

c + h P{D ≤ y} + P (1-P{D ≤ y}) = 0

P {D≤y*} = p-c < 1 o F(y) = p-c

<1
p+h p+h

Mg. Giovana Valverde Ayala 17

Critical Reason

F(y)
1
P-c
P+h

y
y*

Mg. Giovana Valverde Ayala 18

9
For the discrete case

C(y) = E{C(y)}= [cy + hmáx{0, y-D} + pmáx{0,D-y} + A]f(D)dD

Y
E{C(y)} = Cy + h (y - D) f(D) dD + p (D - y) f(D) dD + A
d=0 d=Y

The necessary conditions for optimization are:

E {C(y-1)} ≥ E {C(y)} yE {C(y+1)} ≥ E {C(y)}

Ms. Giovana Valverde Ayala 19

These conditions are sufficient because {C(y)} is a

convex function.
After some algebraic manipulations, the
application of these conditions results in the following
inequalities to determine and *:

P {D≤ y* - 1} ≤ P–C ≤ P {D≤ y*}

p+h

Mg. Giovana Valverde Ayala 20

10
 Problem
The owner of a newsstand wants
determine the number of copies of the newspaper that
must have at the beginning of each day. It costs 30um.
specimen, and the owner sells it for 75um. The sale of the
The newspaper normally occurs at 7:00 am and 2:00 pm.
newspapers that are left at the end of the day are recycled for
a fee of 5um. per copy.
How many copies must the owner have each
tomorrow?
Assuming that daytime demand can be described
like a normal distribution with a mean of 300 copies and
standard deviation of 20 samples.

Ms. Giovana Valverde Ayala 21

Solution
The handling and penalty costs are not defined.
However, the data of the problem tells us that each
The unsold copy will cost the owner 30 - 5 = 25um.
and that the penalty for running out of stock is
75um. per copy.
So we have:
P = 75 um. per copy per day.
h = 25 um. per copy per day.
Now we determine the critical ratio:

P{D ≤ y*} = p - c = 75 - 30 equals 0.45

p+ h 75 +25

Mg. Giovana Valverde Ayala 22

11
According to the equation, the optimal number of newspapers that
the number y* that satisfies must be reserved:

P{D ≤ y*} = 0.45 (1)

As the demand D has a normal distribution with

mean 300 and standard deviation 20 (N(300,20)),
we can normalize equation (1) with respect to the
random variable D and obtain:

P D - 300 ≤ y* - 300) = 0.45

20 20
P ( Z ≤ y* - 300 ) = 0.45(2)
20

Mg. Giovana Valverde Ayala 23

From the standard normal table we have:

P {Z ≤ - 1.25} 0.45 (3)

From (2) and (3) we obtain: y* - 300 = -1.25
20
Y * = 300 – 20 (1.25) = 275

Therefore, the optimal quantity to order is:

y * = 275

Ms. Giovana Valverde Ayala 24

12
 Problem
Bonetón (bookstore) has to decide in November how many calendars
about nature, you must ask for next year.
All calendars that have not been sold by January first will
they go back to the editor, who reimburses 75um per calendar. Bonetón thinks.
that the number of calendars sold by January first follows the
probability distribution shown in table 1. Bonetón wishes to
maximize the expected net utility from the sales of the calendars.
How many calendars should the bookstore order in November?
Table

Mg. Giovana Valverde Ayala 25

Probabilistic Model of a period with Initial stock

VLet's consider that there is an initial stock X

Accessible amount (Y) = initial stock (x) + quantity
produced (y - x)
y ∞
C(y-x) = min {c (y-x) + h [(y - d)] f(d) dD + p y[(d - y)] f(d) dD}
y≥x 0
y ∞
min {cy + h [(y - d)] f(d) dD + p y
[(d - y)] f(d) dD– cx}
y≥x 0

min { cy + L(y) - cx} Therefore:

y≥x

Order (and0– x) if x ≤ y0
Do not order if x > 0y

Mg. Giovana Valverde Ayala 26

13
 Exercise
1.- Consider the one-period model with h = $0.5, p = $4.5 and c =
$0.5. The demand density function is given as:

1/10, 0 ≤ D ≤ 10
f(D) = 0, D > 10

2.- Considering the previous exercise:

a) Determine the order quantity in each of the cases
following:
i) Inventario inicial = 5 unidades.
Initial inventory = 10 units
b) Determine the probability that it does not end
existence during the period.

Mg. Giovana Valverde Ayala 27

Probabilistic Model of a period with stock

Initial and Discrete Demand
Y-1 ∞
E[C(D,y)] = C(y-x) + h (y - d) PD(d) + p (d - y) PD(d)
d=0 d=y
In the discrete case, the necessary and sufficient condition for the
minimum is:
E [C(y-1)] ≥ E [C(y)] y E [C(y+1)] ≥ E [C(y)]
……
E [C(y-1)] = E [C(y)] –C + P – (h+P) P(D≤ y–1)
E [C(y-1)] - E [C(y)] = P – C – (h+P) P(D≤ y–1) ≥ 0

P - C ≥ P(D ≤ y - 1)
P+h
Ms. Giovana Valverde Ayala 28

14
The E [C (y+1)] ≥ E [C (y)] P (D ≤ y) ≥ P - C
P+h
P (D ≤ y0-1) ≤ P - C ≤ P (D ≤ y0)
P+h

Example: h = $1 p = $4 c = $2

D 0 1 2 3 4 5
PD 0.1 0.2 0.25 0.20 0.15 0.10
P{D≤d} 0.1 0.3 0.55 0.75 0.90 1.00

P - C = 4 - 2 = 2 = 0.4 y0= 2
P+h 4+15

Ms. Giovana Valverde Ayala 29

Probabilistic Model of a period with Cost of

Sorting A to K

Let's consider the same previous model but now with

a fixed cost A or K.
Let's assume first that p(.) and h(.)y are linear functions.
∞
L(y) = p (D-y) f(D) dD + h (y-D) f(D) dD
y 0
G(y) = L(y) + Cy
The cost of ordering an amount (y-x) will be:
(1) K + c (y-x) + if L(y)
y > x
(2) L(y) if y ≤ x

Mg. Giovana Valverde Ayala 30

15
 Probabilistic Model of a period with Cost of
Sorting A to K

Cy+L(y)

y
x s x’ S x

Ms. Giovana Valverde Ayala 31

Probabilistic Model of a period with Cost of

Sorting A to K

Find the value of Y that minimizes G(y);

F(S) = (p-c)/(p+h), let's call it the smallest value
what makes that:
Cs+ L (s) = K + cS+ L (S)
In the figure it is evident that:
Six > SK + Cy + L(y) > cx + L(x) for all y > x
K + c(y - x) + L(y) > L(x)
the optimal policy will be not to order if x > S.

Mg. Giovana Valverde Ayala 32

16
 Probabilistic Model of a period with
Sorting Cost A to K
Sis<x’≤S will be:
K + cy + L(y) ≥ cx' + L(x') for all y > x'
K + c(y-x') + L(y) ≥ L(x') the optimal thing will be not to order

Six
Min {K + cy + L(y)} = K + cS + L(S) < cx" + L(x")
y is greater than or equal to x

K + c(S-x”) + L(S) < L(x”) and the optimal policy will be

order up to S, that is to say order (S-x")

Ms. Giovana Valverde Ayala 33

Probabilistic Model of a period with Cost of

Sorting A to K

In general, the optimal policy will be.

If x < sort until S
If x ≥ sno arrange
Where it is held:
F(S) = (p-c)/(p+h) to find S
find the smallest value that makes
Cs + L(s) = K + cS + L(S)

Ms. Giovana Valverde Ayala 34

17
 Example
Find the optimal policy for a horizon of one
period yes :
f(x) = 1/20 0 ≤ X ≤ 20
H=1; p=7; k=10; c=2
What is the optimal amount to order?
FindingS:

F(S) = (p-c) → S=5/8 therefore S= 12.5

(p+h) 20

20 y
L(y) = P (x-y) f(x) dx + h (y-x) f(x) dx
y 0

Mg. Giovana Valverde Ayala 35

Cs+ L(s)= K + cS+ L(S)

2s + 70 - 7s + s²/5 = 10 + 2(12.5) + 70 - 7(12.5) + (12.5)²/5
s² - 25s + 106.5 = 0
therefore = 25 ± √(625 - 426)
2

s = (25 ± 14.1) = (20, 5.45) s= 5.45

2
Resultado :
if x < 5.45 order up to 12.5
if x ≥ 5.45 do not order
Mg. Giovana Valverde Ayala 36

18
 Exercise 1
Find the optimal policy for a horizon of one
period yes :
f(x) = 1/20 for 0 ≤ X ≤ 20

h=1; p=3; k=1.5; c=2

Optimal policy Hallandola (s, S)

Mg. Giovana Valverde Ayala 37

Exercise 2
A rocket spare part becomes obsolete in a year, the cost
the production cost is $2000 and, if it must be produced in
emergency, its cost is $10000, the cost of Inventory
the demand is $100 and follows an exponential distribution
with an average ℷ = 50, there is also a fixed cost of $2000
for production run. Find the optimal policy.

Mg. Giovana Valverde Ayala 38

19
 Exercise 3
A store can buy an item that it will sell for
the Christmas items at a cost of $C/unit and sells them at
$V/unit. V>C; the salvage value of the items
that are not sold in season is S<C. If the function
from the demand is f(x), demonstrate that the optimal value of Y
is given by the solution of:
F(Y*) = V-C
V-S

Ms. Giovana Valverde Ayala 39

20