STOCHASTIC INVENTORY

MODELS
INOPER3 Notes
 Single Period Model (Newsboy
Problem)
 Parameters:
› a = ordering cost
› s = selling price
› c = purchase cost
› v = salvage value
› p = stock-out cost
 Decision Variable
› Q = order quantity
 Random Variable
› x = demand (per period)
› f(x) = demand density function
 Single Period Model (Newsboy
Problem)
 Expected profit = E(P)
 If x < Q:
P1  s * x  v(Q  x)  cQ  a
 If x > Q:
P2  s * Q  p( x  Q)  cQ  a
 Single Period Model (Newsboy
Problem)
 Expected Profit
Q 
E ( P )   [ sx  v(Q  x )  cQ  a] f ( x)dx   [ sQ  p( x  Q)  cQ  a ] f ( x)dx
 Q
Q 
  [ sx  v(Q  x)] f ( x)dx   [ sQ  p ( x  Q)] f ( x)dx  cQ  a
 Q
 Single Period Model (Newsboy
Problem)
 Applying Leibniz Rule:
Q 
dE( P)
 1[sQ  0] f (Q)  0   v( fx)dx  0  1[sQ  0] f (Q)   [s  p] f ( x)dx  c  0
dQ  Q
 
 v[1   f ( x)dx]  (s  p) f ( x)dx  c  0
Q Q

 v  (s  p  v) f ( x)dx  c  0
Q

cv
 f ( x)dx 
s pv
(critical ratio)
Q*
 Single Period Model (Newsboy
Problem)
 If x is discrete, the critical ratio becomes:

cv
 f ( x) 
s pv
Q*

 If B = beginning inventory,
› Q* < B, then order zero (0)
› Q* > B, then order Q* - B
 Example 1
 A small grocery store has the problem of determining the
number of trays of bread to stock each day. A tray costs
Php125 and sells for Php150. The leftover trays at the end of
the day can be sold as day-old bread at Php112.50. The
probability distribution of demand is given below. Find the
optimal order quantity.

x 10 11 12 13 14 15
f(x) 0.15 0.20 0.19 0.18 0.17 0.11
 Example 2

 Given the same problem, find the optimal order

quantity if the demand has the following density
function:
1
f ( x)  ; 10  x  15
5
0 ; elsewhere
 Example 3
 A company is trying to determine the amount of
particular part to manufacture as replacements
before shutting down the production lines of this
model. No future runs will be considered. The
parts sell for Php10, cost Php6 to manufacture and
can be scrapped at Php2 if not sold. The demand
for the product is exponential with a mean of 50
units. At present, there are twenty parts in the
inventory. How many should be produced?
 Example 4

 A book store orders copies of popular magazine

at a cost of Php7.50 per copy. The magazine can
be sold at Php15.00 per copy. At the end of the
month, when the next issue is delivered, any
leftovers can be returned to distributor for a
credit of Php2.50. Monthly demand for the
magazine can be regarded as normal with a mean
of 50 and a standard deviation of 15. How many
copies should the store order?
 Continuous Review Models: Lot
Size, Reorder Point Models
 Back order Case
 Stock out Case
 Back Order Case
 Parameters
› a = fixed ordering cost
› h = holding cost per unit
› p = back-order cost per unit
› d = expected demand per planning horizon
 Decision Variable
› Q = order quantity
› r = reorder point
 Random Variable
› x = demand during lead time
› f(x) = density function of demand during lead time
› µ = mean demand during lead time
Back Order Case

Inventory

Time
 Back Order Case

 EAC (Q, r) = Total Expected Annual Cost

= ordering cost + holding cost +
back order cost
= OC + HC + BC
 Ordering cost:
d
OC  a
Q
 Back Order Case

 Holding cost:
HC  h * I ave

I max  I min
I ave 
2
› Expected inventory level just before arrival of Q = r -
µ (can be negative)
› Expected inventory level just after arrival of Q = Q + r
-µ
 Back Order Case

 Holding cost (cont.) :

Qrr Q
I ave   r
2 2

Q 
HC  h   r   
2 
 Back order cost:
› Let B(r) = number of back ordered units
› B(r) = 0 ; x < r
› B(r) = x - r ; x > r
 Back Order Case

 Back order cost (cont.):

r 
B(r )   0 * f ( x)dx   ( x  r ) f ( x)dx
 r

  ( x  r ) f ( x)dx
r

d
BC  p * B(r )
Q
 Back Order Case

 Total Expected Annual Cost (EAC):

ad Q pd
EAC (Q, r )   h[  r   ]  B(r )
Q 2 Q

 Applying partial derivative:

EAC ad h pd
   2 B (r )  0
Q Q 2 Q

* 2d [a  pB(r )
Q  Equation (1)
h
 Back Order Case

 Applying partial derivative:

EAC pd
 0h B' (r )  0
r Q


› Where: B(r )   ( x  r ) f ( x)dx

r
› Applying Leibniz Rule:

B' (r )  0  (1)(r  r ) f (r )   (1) f ( x)dx
r

B ' (r )    f ( x)dx
r
 Back Order Case
 (cont.) EAC pd


r
 h
Q  f ( x)dx  0
r


hQ
r
f ( x)dx 
pd Equation (2)

 Iterative Process:
1. Set B(r) = 0 and solve for Q1 using Equation (1)
2. Solve r1 using Equation (2)
3. Solve B(r1)
4. Solve Q2
5. Solve r2
6. Continue process until vales of Q and r converge.
 Back Order Case

 Note: When f(x) is complicated, the evaluation of

B(r) may not be trivial.
 When x is normal:
r r
B(r )  f      r G  
     
 Where: r
f  is the value of the function at (r- µ)/σ
  

r
G  is the area to the right of z
  
2
1  r  
r 1  
2   
f  e
   2
 Example 5

 Items can be ordered at a cost of Php20. It costs

Php1 per week to store one unit and Php5 is the
cost incurred for every item back-ordered. Lead
time equals one week. Find the optimal order
quantity and reorder point assuming:
(a) demand has the following distribution
X 10 11 12 13 14
f(x) 0.3 0.25 0.2 0.15 0.1

(b) demand is uniformly distributed

1
f ( x)  ; 10  x  14
4
 Example 6
 A department sells blank tape under its own brand
name. Because of special labeling and packaging,
there is a lead time of 5.2 weeks. Assume that the
demand during lead time is normal with a mean of
1000 tapes and standard deviation of 250. The cost
of paper work and handling associated with placing
an order is Php100 per order. Holding cost is
Php0.5 per tape per year. The penalty is Php1 per
tape back-ordered. Find the optimal quantity and
reorder point.
 Stock-out Case
 Parameters
› a = fixed ordering cost
› h = holding cost
› p = stock-out cost per unit
› d = expected during per planning horizon
 Decision Variable
› Q = order quantity
› r = reorder point
 Random Variable
› x = demand lead time
› f(x) = density function of demand during lead time
› µ = mean demand during lead time
 Stock-out Case

 EAC (Q, r) = Expected Total Annual Cost

= ordering cost + holding cost +
stock-out cost
= OC + HC + SC
 Ordering cost:
d
OC  a
Q
 Holding cost: HC  h * I ave
Q 
HC  h   r    B (r )
2 
 Stock-out Case

 Stock-out cost:
› Let B(r) = number of stocked out units

B (r )   ( x  r ) f ( x)dx
r

d
SC  p * B(r )
Q

 Total Expected Annual

ad Q Cost pd
EAC(Q, r )   h[  r    B(r )]  B(r )
Q 2 Q
 Stock-out Case
 Applying partial derivative:

2d [a  pB(r ) Equation (1)

Q* 
h

hQ

r
f ( x)dx 
pd  hQ
Equation (2)

 Iterative Process:
1. Set B(r) = 0 and solve for Q1 using Equation (1)
2. Solve r1 using Equation (2)
3. Solve B(r1)
4. Solve Q2
5. Solve r2
6. Continue process until vales of Q and r converge.
 Stock-out Case

 Note: When f(x) is complicated, the evaluation of

B(r) may not be trivial.
 When x is normal:
r r
B(r )  f      r G 
     
 Where: r
f  is the value of the function at (r- µ)/σ
  
r
G  is the area to the right of z
  
2
1  r  
r 1  
2   
f  e
   2
 Example 7

 Consider Example 5. Assume Stock out case.

 Consider Example 6. Assume Stock out case.
Periodic Review Model for
Stochastic Demand

m-μ

tr L
 Periodic Review Model for
Stochastic Demand
 Let
› tr = elapsed time between reviews
› m = target level to which we want to restore the inventory
› Ir = inventory at the time of review
› f(x) = density function of demand during lead time
› a = combined order cost and cost of reviewing
› h = holding cost per unit per T
› p = back order cost
› μ = mean of density function of demand during lead time
› d = expected value of demand
› Beginning inventory of a cycle = m – μ
› Expected Demand during tr = d * tr
› Ending Inventory = m – μ – d * tr
 Periodic Review Model for
Stochastic Demand
a
 Total Order and Review Cost =
tr
 Total Holding Cost =
 m    m    dt r   1 
H   H  m    dt r 
 2   2 
 Total time when demand is
present = tr + L
 f(D/tr+L) = density function of demand given a
time period of tr + L
 Periodic Review Model for
Stochastic Demand
 Bm = expected back ordered units

Bm   ( D  m) f ( D / t r  L)dD
m
p
Bm
 Expected back ordered cost = tr
 Therefore,
a  1  p 
E (c )   H m    dt r    ( D  m) f ( D / t r  L)dD
tr  2  tr m
 Using Leibniz Rule: dE (c) p 
 H   f ( D / t r  L)dD  0
dm tr m
 Ht r
 m
f ( D / t r  L)dD 
p
 Example 8

 Assume that an item can be ordered at a cost of

Php20. It costs Php1 per week to store it and
Php5 is incurred for every item back ordered.
The lead time is 1 week. A Php10 review cost is
also incurred. Assume that the demand per week
is normal with a mean of 12 and a variance of 1.
Find the target level of inventory m of the review
period is 2 weeks.
 Solution to Example 8

 Given: tr = 2 weeks
L = 1 week
a = 20 + 10 = 30
h = 1 per unit per week
p = 5 per unit
Normal distribution of demand with μ = 12 and σ2 = 1
tr + L = 3 weeks
 Solution to Example 8

 f(D/3) = normal with μT = 36 and σ2T = 3

 1(2)

m
f ( D / 3)dD 
5
 0.4

P( D  m)  0.4
m  36
z 0.6  0.255 
3
m  36.44

 36.44  36    36.44  36 
Bm  3 f    G 36  36.44  0.4929
 3    3 

30  1  5
E (c )   136.44  12  (12)(2)  (0.4929)  28.67
2  2  2
Solution to Example 8

tr E(c)
1 37.98
2 28.67
2.2 28.45
2.3 28.41
2.4 28.43
3.0 29.29