Inventory Models For Probabilistic Demand:: Cost & Service Trade-Offs

CTL.
SC1x -Supply Chain & Logistics Fundamentals
Inventory Models for

Probabilistic Demand:
Cost & Service Trade-offs
MIT Center for

Transportation & Logistics
Items to Cover
•  Comparing Inventory Performance Metrics

•  Inputted versus Implied Metrics
•  Periodic Review Policies
•  Example Problem: ShopCo
•  Trading off Lead Time and Review Period
CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Cost & Service Trade-offs 2
Notation µDL , σDL = Expected and Standard
Deviation of Demand over Lead Time
D = Average Demand (units/time) (units)
c = Variable (Purchase) Cost ($/unit) µDL+R , σDL+R= Expected and Standard
h = Carrying or Holding Charge ($/ Deviation of Demand over Lead Time
inventory $/time) plus Review Period (units)
ct = Fixed Ordering Cost ($/order) k = Safety Factor
ce = c*h = Excess Holding Cost ($/unit/
time) s = Reorder point (units)
cs = Shortage Cost ($/unit/time) S = Order up to Point (units)
Q = Replenishment Order Quantity R = Review Period (time)
(units/order)
L = Replenishment Lead Time (time)
IFR = Item Fill Rate (%)
T = Order Cycle Time (time/order)
CSL = Cycle Service Level (%)
N = 1/T = Orders per Time (order/time)
CSOE = Cost of Stock Out Event ($/event)
IP = Inventory Position (units)
CSI = Cost per Item Short
IOH = Inventory on Hand (units)
IOO = Inventory On Order (units)
E[US] = Expected Units Short (units)
G(k) = Unit Normal Loss Function
Comparing Inventory Performance Metrics
Inventory Performance Metrics
•  Establishes Safety Stock
•  Finds k value
s = µ DL + kσ DL
•  Expected Cost of Safety Stock = cekσDL
•  Service Based Metrics – set k to meet expected LOS
•  Cycle Service Level (CSL)
•  Probability of not stocking out during cycle CSL = P "# x ≤ k $%
•  Item Fill Rate (IFR)
•  Expected percentage of demand met during each cycle σ DLG[k]
IFR = 1−
Q
•  Cost Based Metrics – find k that minimizes total costs
•  Cost per Stockout Event (CSOE) & D)
•  Penalty of B1 if any stock out occurs ! # ! #
E "CSOE $ = (B1 )P " x ≥ k $( +
'Q*
•  Cost per Item Short (CIS)
•  Penalty of cs for each item short per cycle % D(
! #
E "CIS $ = csσ DLG(k) ' *
&Q)
Performance Metrics v. Safety Stock Costs
100%
$325,000
95%
$300,000
90%
•  Demand ~N(62,000, 8,000) $275,000
•  ce = 15 $/unit-year
Total Annual Shortage Cost (CSOE or CIS)

$250,000
85% •  Q*= 5,200 units/order
Level of Service % (CSL or IFR)
•  L = 2 weeks $225,000
80% •  B1= 50,000 $/event $200,000

•  cs= 45 $/unit-year
75% •  µDL = 2,385 units $175,000
•  σDL = 1,569 units $150,000

70%
$125,000
65% $100,000
$75,000
60%
$50,000
55%
$25,000
50% $-
$0 $5,000 $10,000 $15,000 $20,000 $25,000 $30,000 $35,000 $40,000 $45,000 $50,000 $55,000 $60,000 $65,000 $70,000
Expected Annual Cost of Safety Stock
E !"Cost of SS#$ = ce kσ DL
CSL IFR TotCost_CSOE TotCost_CIS
Lead Time v. Safety Stock Costs
100%
95%
Level of Service % (CSL)
90%
85%
80%
75% CSL_LT=1 wk
70%
CSL_LT=2 wk
65%
CSL_LT=4 wk
60%
55% CSL_LT=8 wk
50%
$- $10,000 $20,000 $30,000 $40,000 $50,000 $60,000 $70,000 $80,000 $90,000 $100,000
100%
Level of Service % (IFR)
98%
96%
94% IFR_LT=1 wk
IFR_LT=2 wk
92%
IFR_LT=4 wk
90% IFR_LT=8 wk
88%
$- $10,000 $20,000 $30,000 $40,000 $50,000 $60,000 $70,000 $80,000 $90,000 $100,000
Inputted vs. Implied Metrics
Safety Stock Logic E !"Cost of SS#$ = ce kσ DL
Re-order
Safety
CSL Point TRC(s,Q)
Stock
CSL = P "# x ≤ k $%
SS=kσDL s=µDL+kσDL
P[SO]=1-CSL
()
E !"US #$ = σ DLG k
P[SO] k G(k) E[US]
Qc ! $
P !" SO#$ = e B1 D
Dcs k = 2ln #
# c Qσ
&
&
G(k)=(1-IFR)Q/σDL
" e DL 2π %
CIS CSOE CIS

IFR
(cs) (B1) (cs)
Figure adapted from Dr. Jim Master’s ESD.260 Course Notes (2002)
Periodic Review Policies
Assumptions: Periodic Review Policies
•  Demand •  Discounts
n  Constant vs Variable n  None
n  Known vs Random n  All Units vs Incremental vs One Time
n  Continuous vs Discrete •  Excess Demand
•  Lead Time n  None
n  Instantaneous n  All orders are backordered
n  Constant vs Variable n  Lost orders
n  Deterministic vs Stochastic n  Substitution
n  Internally Replenished •  Perishability
•  Dependence of Items n  None
n  Independent n  Uniform with time
n  Correlated n  Non-linear with time
n  Indentured •  Planning Horizon
•  Review Time n  Single Period
n  Continuous vs Periodic n  Finite Period
•  Number of Locations n  Infinite
n  One vs Multi vs Multi-Echelon •  Number of Items
•  Capacity / Resources n  One vs Many
n  Unlimited •  Form of Product
n  Limited / Constrained n  Single Stage
n  Multi-Stage
Periodic Review Policies
•  Order-Up-To-Level (R, S) •  Hybrid (R, s, S) System
n  Policy: Order S-IP every n  Policy: Every R time periods,
R time periods Order S-IP if IP ≤ s,
n  Replenishment cycle system if IP>s then do not order
n  General case for many policies
S S
Inventory Position
Inventory Position
L L Time
L L Time
R R R R R
Notation
s = Reorder Point S = Order-up-to Level L = Replenishment Lead Time
Q = Order Quantity R = Review Period IOH= Inventory on Hand
IP = Inventory Position = (IOH) + (Inventory On Order) – (Backorders)
Periodic Review Policies (R, S)
S
Inventory Position
L L Time
R R
Differences from Continuous Review Policy (s, Q)

•  How much to order?
•  How long should safety stock cover?
Periodic vs Continuous Review
•  Convenient transformation of (s, Q) to (R, S)
n  (s, Q)= Continuous, order Q when IP≤s
n  (R, S)= Periodic, order up to S every R time periods
•  Allows for the use of all previous (s, Q) decision rules
n  s for continuous system becomes S for periodic system
n  Q for continuous system becomes D*R for periodic system
n  L for a continuous system becomes R+L for periodic system
•  Approach
(s, Q) (R, S)
n  Make transformations
n  Solve for (s, Q) using transformations s ó S
n  Determine final policy, so that
Q ó D*R
S = xDL+R + kσDL+R L ó R+L
Example: ShopCo
CTL.SC1x - Supply Chain and Logistics Fundamentals Lesson: Cost & Service Trade-offs
Example: ShopCo
•  Background:
n  ShopCo is a North America based large store
format retailer of home improvement products
with >2,000 stores. Each ShopCo store generally operates independently:
ordering and receiving product directly from its suppliers.
n  One supplier (Hurricane Drills) sells a portfolio of electric drills that, on average,
cost ShopCo $75 each. Each store uses periodic review policies to order
directly from Hurricane and uses an annual holding charge of 15%. Assume 52
week year.
•  Problem:
n  Find the (R, S) ordering policy for Hurricane drills for store #1301 given:
w  Forecasted annual demand of Hurricane drills is ~N(3,400, 400)
w  Lead Time is 1 week
w  Review Period is 4 weeks
w  Desired CSL = 95%
w  Hurricane has a minimum order quantity (MOQ) of 240 drills
w  Orders need to be in multiples of 12 drills to fit on pallets
n  What is the expected annual cost of cycle and safety stock?
Case adapted from Anand, S. and Song, X. (2011) “Supply Chain Responsiveness for a Large Retailer,” MIT Supply Chain Management Program Thesis.
Image Source:http://commons.wikimedia.org/wiki/File:Hardware_Store.jpg
Example: ShopCo
•  Finding Order Policy:
n  Find Q = D*R = (3,400 units/year)(4/52 years) = 261.5 ≈ 264 units (why?)
n  Find R+L = 4 weeks + 1 week = 5 weeks or 0.0962 years
so that, n= 52/5 = 10.4 “coverage” periods per year
n  Find µDL+R = (3,400)/(10.4) = 326.9 ≈ 327 units
n  Find σDL+R = (400)/(√10.4) = 124.03 ≈124 units
n  Find k where CSL = 0.95 or P[x≤k] = 0.95, k=1.644 = 1.64
n  Find S = µDL+R + kσDL+R = 327 + (1.64)(124) = 530.4 ≈ 530 units
Policy: Order up to 530 units every 4 weeks.
•  Finding Cost of Cycle & Safety Stock:

n  Cost of Cycle Stock = ce(DR/2) = (75)(0.15)(264/2) = $1,485 per year
n  Cost of Safety Stock = cekσDL+R = (75)(0.15)(1.64)(124) = $2,288 per year
Trading Off Lead Time and Review
Period
Example: ShopCo continued
•  New Mixing Center Strategy:
n  ShopCo has decided to deploy a fulfillment strategy
where each store orders from its Regional Distribution Center (RDC), instead
of directly to the supplier.
n  Each ShopCo RDC then consolidates orders from its dedicated stores and
places a combined order to the vendor. The vendor will then ship to each of
the RDCs where ShopCo “mixes” the products from multiple suppliers to
distribute a single combined load to each store.
Case adapted from Anand, S. and Song, X. (2011) “Supply Chain Responsiveness for a Large Retailer,” MIT Supply Chain Management Program Thesis.
Image Sources: http://commons.wikimedia.org/wiki/File:Hardware_Store.jpg and Anand & Song (2011)
•  Revised Problem with Mixing Centers:
n  Find the (R, S) ordering policy for Hurricanes for store #1301 given:
w  Forecasted annual demand of Hurricane drills is ~N(3,400, 400)
w  Desired CSL = 95%
w  Lead Time is now 10 days (call this 1.5 weeks for simplicity)
w  Review Period is reduced to 2 weeks
w  ShopCo’s RDCs do not have a minimum order quantity (MOQ) to stores (why?)
w  Orders still need to be in multiples of 12 drills to fit on pallets (why?)
n  What is the expected annual cost of cycle and safety stock?
Example: ShopCo with Mixing Strategy
•  Finding Order Policy:
n  Find Q = D*R = (3,400 units/year)(2/52 years) = 130.8 ≈ 132 units (why?)
n  Find R+L = 2 weeks + 1.5 week = 3.5 weeks or 0.0673 years
so that, n= 52/3.5 = 14.86 “coverage” periods per year
n  Find µDL+R = (3,400)/(14.86) = 228.8 ≈ 229 units
n  Find σDL+R = (400)/(√14.86) = 103.76 ≈ 104 units
n  Find k where CSL = 0.95 or P[x≤k] = 0.95, k=1.644 = 1.64
n  Find S = µDL+R + kσDL+R = 229+ (1.64)(104) = 399.56 ≈ 400 units
Policy: Order up to 400 units every 2 weeks.
•  Finding Cost of Cycle & Safety Stock:

n  Cost of Cycle Stock = ce(DR/2) = (75)(0.15)(132/2) ≈ $743 per year
n  Cost of Safety Stock = cekσDL+R = (75)(0.15)(1.64)(104) ≈ $1,919 per year
Strategy Lead Review Cycle Safety Avg. Inventory

Time Period Stock Stock Costs
(weeks) (weeks) ($/year) ($/year) ($/year)
Direct-to-Store 1 4 1,485 2,288 3,773
Mixing Centers 1.5 2 743 1,919 2,662
Which is better? Which did the store managers prefer?
Relationship Between L & R
•  Average Inventory Costs = ce[DR/2 + kσDL+R + LD]
•  Individual Impacts
n  Increasing Lead Time L
è Increases Safety Stock non-linearly
è Increases Pipeline Stock linearly
n  Increasing Review Period, R:
è Increases Safety Stock non-linearly
è Increases Cycle Stock linearly
•  Combined Impacts
n  Can be used to trade Replenishment speed (L) for frequency (R)
n  Determine which is the right mix
Key Points from Lesson
Key Points
•  Inventory Performance Metrics

•  Service Based: IFR vs. CSL
•  Cost Based: CSOE vs. CIS
•  Inputted vs. Implied Metrics
•  Designing to one metric sets the others
•  Can backwards calculate implied values
•  Periodic Review (R, S)
•  Very commonly used
•  Use the (s, Q) rules with simple transformations
Q è D*R, sèS, LèL+R
•  Changing L & R have different impacts on inventory
CTL.SC1x -Supply Chain & Logistics Fundamentals
Questions, Comments, Suggestions?

Use the Discussion!
MIT Center for

Transportation & Logistics caplice@mit.edu
continuous Review and Periodic Review
These terms refer to the frequency of review to determine when orders must be placed for
replenishment.
In the continuous review process, the inventory levels are continuously reviewed, and as soon as
the stocks fall below a pre-determined level (usually called, reorder point, or reorder level),
replenishment order is placed. As more and more companies start using sophisticated IT systems
to track their inventories in real-time, the continuous review method becomes a viable and
optimal way to plan for replenishment.
Under periodic review, the inventory levels are reviewed at a set frequency. At the time of
review, if the stock levels are below the pre-determined level, then an order for replenishment is
placed, otherwise it is ignored till the next cycle. This method provides a viable process
alternative to the continuous review by segmenting the merchandise into review buckets. This
makes it easier to manage when the process is manual, or the number of items involved is
extremely large, or when constraints on ordering-day exist.
Order Quantity and Order up-to Level
These terms refer to the process that is used to determine how much is ordered when a
replenishment order is placed.
In the first process, the “order quantity” is fixed. If the review determines that an order should be
placed, then the order for a pre-defined quantity for that item-location combination is placed for
replenishment. The order quantity for all replenishment orders is fixed in this method, though
order day may vary or may be fixed depending on the review method.
The second process defines a pre-determined “order up-to level” instead. The actual order
quantity is determined as the difference between the on-hand stock on the review day, and the
pre-determined “order up-to level”. The order quantity in this process will differ from one order
to another depending on the on-hand quantity on the day of the review.
Between these two sets of parameters, four basic reordering process options become available.
Options for Re-ordering Process
Based on the above two parameters, the reordering process can be deployed in the four basic
ways. The diagrams below depict these variations of the process.
• Modeling of the Time Axis
Inventory policies differ in two aspects, namely the mechanism used to trigger
replenishment orders and the decision rule that specifies the determination of the
order size. The specific inventory policies are defined through the combination of
the decision variables s (reorder point), r (review interval, order cycle), q (order
quantity) and S (order level) as follows:
• (s,q) policy,
• (r,S) policy,
• (s,S) policy.
(s,q) policy
Under the (s,q) policy, the point in time at which replenishment orders are
triggered, depends on the size of the reorder point s, whereas the order quantity
q is constant over time. In the ideal (textbook) form of the (s,q) policy, the
inventory position is continuously monitored. The inventory position is the sum of
the inventory on hand plus the inventory on order minus the outstanding
backorders (backlog). The inventory management system (or the inventory
manager) acts according to the following decision rule: If at a review instant the
inventory position has reached the reorder point s (from above), then launch a
replenishment order of size q.
In reality the inventory is not monitored continuously. In contrast, the

replenishment decisions are made in discrete time intervals, usually at the end of a
day. In addition, often demand sizes are greater than one unit. Under these
conditions, the analysis of the (s,q) policy as presented in many textbooks in false,
as the so-called undershoot is neglected. In the above figure, the undershoot is the
difference between s and the inventory position at the moment immediately before
a new replenishment order is released. Negleting the undershoot usually results in
significant over-estimation of the service level (under-estimation of the required
safety stock).
(r,S) policy
If an (r,S) inventory policy is in effect, the points in time at which replenishment
orders are released are determined through the review interval r. The inventory
management system proceeds according to the following decision rule: In constant
intervals of r periods launch a replenishment order that raises the inventory
position to the target order level S. Obviously, the (r,S) policy is an inventory
policy with periodic review. The order size at a time of a review depends on the
demands and the development of the inventory observed in the preceding periods.
If r=1, then this poliy is called base-stock policy.
(s,S) policy
Under an (s,S) inventory policy, the points in time when an order is triggered are
determined policy, i. e. through the reorder point s. However, the order quantity is
now, similar to the (r,S) policy, a function of the inventory development over time.
In the literature this policy is sometimes characterized with the help of a third
parameter which specifies the length of the review interval r. In this notation the
policy is called (r,s,S) policy. In the case of r=0, continuous review is in effect. If
demands arrive unit-sized, then the (r=0,s,S) policy is identical to the (s,q) policy
with continuous review.
For the determination of the optimum safety stock under conditions of uncertainty
the demand during the risk period plays a central role.
The risk period is composed of
• the review period and

• the replenishment lead time.
Stochastic demand occurs within this time span that usually comprises several
periods. In order to compute the parameters of an inventory policy, we must know
the probability distribution of the demand during the risk period.
• Demand during the Risk Period

Inventory Models For Probabilistic Demand:: Cost & Service Trade-Offs

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Inventory Models For Probabilistic Demand:: Cost & Service Trade-Offs

Uploaded by

Copyright:

Available Formats

CTL.

SC1x -Supply Chain & Logistics Fundamentals

Inventory Models for

MIT Center for

• Comparing Inventory Performance Metrics

Total Annual Shortage Cost (CSOE or CIS)

80% • B1= 50,000 $/event $200,000

• σDL = 1,569 units $150,000

P[SO] k G(k) E[US]

CIS CSOE CIS

IP = Inventory Position = (IOH) + (Inventory On Order) – (Backorders)

Differences from Continuous Review Policy (s, Q)

S = xDL+R + kσDL+R L ó R+L

Policy: Order up to 530 units every 4 weeks.

• Finding Cost of Cycle & Safety Stock:

Policy: Order up to 400 units every 2 weeks.

• Finding Cost of Cycle & Safety Stock:

Strategy Lead Review Cycle Safety Avg. Inventory

Which is better? Which did the store managers prefer?

• Inventory Performance Metrics

Questions, Comments, Suggestions?

MIT Center for

Order Quantity and Order up-to Level

Options for Re-ordering Process

In reality the inventory is not monitored continuously. In contrast, the

The risk period is composed of

• the review period and

• Demand during the Risk Period

You might also like

•  Comparing Inventory Performance Metrics

80% •  B1= 50,000 $/event $200,000

•  σDL = 1,569 units $150,000

•  Finding Cost of Cycle & Safety Stock:

•  Finding Cost of Cycle & Safety Stock:

•  Inventory Performance Metrics