Transcript Deterministic Inventory Management
INVENTORY MANAGEMENT
Operations Management Dr. Ron Lembke
Purposes of Inventory
Meet anticipated demand Demand variability Supply variability Decouple production & distribution permits constant production quantities Take advantage of quantity discounts Hedge against price increases Protect against shortages
2006 2007 13.81
1857 24.0% 446 801 58 1305 9.9
US Inventory, GDP ($B)
14,000 12,000 10,000 8,000 6,000 4,000 2,000 19 84 19 86 19 88 19 90 19 92 19 94 19 96 Business Inventories 19 98 20 00 US GDP 20 02 20 04
US Inventories as % of GDP
25.0% 20.0% 15.0% 10.0% 5.0% 0.0% 1984 1986 1988 1990 1992 1994
Year
1996 1998 2000 2002 2004 Source: CSCMP, Bureau of Economic Analysis
Two Questions
Two main Inventory Questions: How much to buy?
When is it time to buy?
Also: Which products to buy?
From whom?
Types of Inventory
Raw Materials Subcomponents Work in progress (WIP) Finished products Defectives Returns
Inventory Costs
What costs do we experience because we carry inventory?
Inventory Costs
Costs associated with inventory: Cost of the products Cost of ordering Cost of hanging onto it Cost of having too much / disposal Cost of not having enough (shortage)
Shrinkage Costs
How much is stolen?
2% for discount, dept. stores, hardware, convenience, sporting goods 3% for toys & hobbies 1.5% for all else Where does the missing stuff go?
Employees: 44.5% Shoplifters: 32.7% Administrative / paperwork error: 17.5% Vendor fraud: 5.1%
Inventory Holding Costs
Category Housing (building) cost Material handling Labor cost Opportunity/investment Pilferage/scrap/obsolescence Total Holding Cost % of Value 4% 3% 3% 9% 2% 21%
Inventory Models
Fixed order quantity models How much always same, when changes Economic order quantity Production order quantity Quantity discount Fixed order period models How much changes, when always same
Economic Order Quantity
Assumptions Demand rate is known and constant No order lead time Shortages are not allowed Costs: S - setup cost per order H - holding cost per unit time
EOQ
Inventory Level Q*
Optimal Order Quantity Decrease Due to Constant Demand
Time
EOQ
Inventory Level Q*
Optimal Order Quantity Instantaneous Receipt of Optimal Order Quantity
Time
EOQ
Q* Inventory Level Reorder Point (ROP) Lead Time Time
EOQ
Q* Inventory Level Reorder Point (ROP) Lead Time
Average Inventory Q/2
Time
Total Costs
Average Inventory = Q/2 Annual Holding costs = H * Q/2 # Orders per year = D / Q Annual Ordering Costs = S * D/Q Cost of Goods = D * C Annual Total Costs = Holding + Ordering + CoG
TC
(
Q
)
H
*
Q
2
D S
*
Q
C
*
D
How Much to Order?
Annual Cost Holding Cost = H * Q/2 Order Quantity
How Much to Order?
Annual Cost Ordering Cost = S * D/Q Holding Cost = H * Q/2 Order Quantity
How Much to Order?
Annual Cost Total Cost = Holding + Ordering Order Quantity
How Much to Order?
Annual Cost Total Cost = Holding + Ordering Optimal Q Order Quantity
Optimal Quantity
Total Costs = Take derivative with respect to Q = Solve for Q:
H
2
DS Q
2
Q
2
H
*
Q
2
S D
*
Q
C
*
D H
2
S D
*
Q
2 0
Set equal to zero
2
DS H Q
2
DS H
Adding Lead Time
Use same order size
Q
Order before inventory depleted 2
DS H
d
= average demand rate (per day) L = lead time (in days) both in same time period (wks, months, etc.)
A Question:
If the EOQ is based on so many horrible assumptions that are never really true, why is it the most commonly used ordering policy?
Profit function is very shallow Even if conditions don’t hold perfectly, profits are close to optimal Estimated parameters will not throw you off very far
Quantity Discounts
How does this all change if price changes depending on order size?
Holding cost as function of cost: H = I * C Explicitly consider price:
Q
2
DS I
C
Discount Example
D = 10,000 S = $20 I = 20% PriceQuantity EOQ c = 5.00
Q < 500 4.50
501-999 3.90
Q >= 1000 633 666 716
Discount Pricing
Total Cost
Price 1 Price 2 Price 3 500
X
633
X
666
X
716 1,000
Order Size
Discount Pricing
Total Cost
Price 1 Price 2 Price 3 500
X
633
X
666
X
716 1,000
Order Size
Discount Example
Order 666 at a time: Hold 666/2 * 4.50 * 0.2= $299.70
Order 10,000/666 * 20 = $300.00
Mat’l 10,000*4.50 = $45,000.00 45,599.70
Order 1,000 at a time: Hold 1,000/2 * 3.90 * 0.2=$390.00
Order 10,000/1,000 * 20 = $200.00
Mat’l 10,000*3.90 = $39,000.00
39,590.00
Discount Model
1. Compute EOQ for next cheapest price 2. Is EOQ feasible? (is EOQ in range?) If EOQ is too small, use lowest possible Q to get price.
3. Compute total cost for this quantity 4.
5.
Repeat until EOQ is feasible or too big.
Select quantity/price with lowest total cost.
INVENTORY MANAGEMENT -- RANDOM DEMAND
Random Demand
Don’t know how many we will sell Sales will differ by period Average always remains the same Standard deviation remains constant
Impact of Random Demand
How would our policies change?
How would our order quantity change?
How would our reorder point change?
Mac’s Decision
How many papers to buy?
Average = 90, st dev = 10 Cost = 0.20, Sales Price = 0.50
Salvage = 0.00
Cost of overestimating Demand, C O C O = 0.20 - 0.00 = 0.20
Cost of Underestimating Demand, C U C U = 0.50 - 0.20 = 0.30
Optimal Policy
G(x) = Probability demand <= x Optimal quantity: Pr(D Q)
C o C
u C u
Mac: G(x) = 0.3 / (0.2 + 0.3) = 0.6
From standard normal table, z = 0.253
=Normsinv(0.6) = 0.253
Q* = avg + z s = 90+ 2.53*10 = 90 +2.53 = 93
Optimal Policy
If units are discrete, when in doubt, round up If u units are on hand, order Q - u units Model is called “newsboy problem,” newspaper purchasing decision By time realize sales are good, no time to order more By time realize sales are bad, too late, you’re stuck Similar to the problem of # of Earth Day shirts to make, lbs. of Valentine’s candy to buy, green beer, Christmas trees, toys for Christmas, etc., etc.
Random Demand – Fixed Order Quantity If we want to satisfy all of the demand 95% of the time, how many standard deviations above the mean should the inventory level be?
Probabilistic Models
Safety stock = x From statistics,
m
z
x
s m
L
Safety stock Therefore, z =
s
L From normal table z .95
= 1.65
& Safety stock = z
s
L Safety stock = z
s
L R =
m
= 1.65*10 = 16.5
+ Safety Stock =350+16.5 = 366.5 ≈ 367
Random Example
What should our reorder point be?
demand over the lead time is 50 units, with standard deviation of 20 want to satisfy all demand 90% of the time (i.e., 90% chance we do not run out) To satisfy 90% of the demand, z = 1.28
Safety stock = z σ L = 1.28 * 20 = 25.6
R = 50 + 25.6 = 75.6
St Dev Over Lead Time
What if we only know the average daily demand, and the standard deviation of daily demand?
Lead time = 4 days, daily demand = 10, standard deviation = 5, What should our reorder point be, if z = 3?
St Dev Over LT
If the average each day is 10, and the lead time is 4 days, then the average demand over the lead time must be 40.
d
*
L
10 * 4 40 What is the standard deviation of demand over the lead time?
Std. Dev. ≠ 5 * 4
St Dev Over Lead Time
Standard deviation of demand = s
L
L days
s
day
4 5 10
R
d
*
L
z
s
L
R = 40 + 3 * 10 = 70
d
*
L
z L days
s
day
Service Level Criteria
Type I: specify probability that you do not run out during the lead time Probability that 100% of customers go home happy Type II: proportion of demands met from stock Percentage that go home happy, on average Fill Rate: easier to observe, is commonly used G(z)= expected value of shortage, given z. Not frequently listed in tables
G
(
z
) s
Q L
1
Fill Rate
Two Types of Service
Cycle Demand 1 180 2 3 4 75 235 140 10 Sum 5 6 7 8 9 180 200 150 90 160 40 1,450 Stock-Outs 0 0 45 0 0 10 0 0 0 0 55
Type I:
8 of 10 periods 80% service
Type II:
1,395 / 1,450 = 96%
FIXED-TIME PERIOD MODELS
Fixed-Time Period Model
Every T periods, we look at inventory on hand and place an order Lead time still is L.
Order quantity will be different, depending on demand
Fixed-Time Period Model: When to Order?
Inventory Level Target maximum Time Period
Fixed-Time Period Model: : When to Order?
Inventory Level Target maximum Period Period Time
Fixed-Time Period Model: When to Order?
Inventory Level Target maximum Period Period Time
Fixed-Time Period Model: When to Order?
Inventory Level Target maximum Period Period Period Time
Fixed-Time Period Model: When to Order?
Inventory Level Target maximum Period Period Period Time
Fixed-Time Period Model: When to Order?
Inventory Level Target maximum Period Period Period Time
Fixed Order Period
Standard deviation of demand over T+L = s
T
L
T
L
s T = Review period length (in days) σ = std dev per day Order quantity (12.11) =
q
d
(
T
L
)
z
s
T
L
I
Inventory Recordkeeping
Two ways to order inventory: Keep track of how many delivered, sold Go out and count it every so often If keeping records, still need to double-check Annual physical inventory, or Cycle Counting
Cycle Counting
Physically counting a sample of total inventory on a regular basis Used often with ABC classification A items counted most often (e.g., daily) Advantages Eliminates annual shut-down for physical inventory count Improves inventory accuracy Allows causes of errors to be identified
Fixed-Period Model
Answers how much to order Orders placed at fixed intervals Inventory brought up to target amount Amount ordered varies No continuous inventory count Possibility of stockout between intervals Useful when vendors visit routinely Example: P&G rep. calls every 2 weeks
ABC Analysis
Divides on-hand inventory into 3 classes A class, B class, C class Basis is usually annual $ volume $ volume = Annual demand x Unit cost Policies based on ABC analysis Develop class A suppliers more Give tighter physical control of A items Forecast A items more carefully
Classifying Items as ABC
% Annual $ Volume 100 80 60 40 20 0 A B 0 50
Items A B C %$Vol %Items 80 15 5 15 30 55
100 C % of Inventory Items 150
ABC Classification Solution
Stock # Vol. Cost $ Vol. % 206 105 019 144 207 Total 26,000 $ 36 $936,000 71.1 200 2,000 600 55 120,000 110,000 9.1 8.4 20,000 7,000 4 10 80,000 70,000 6.1 5.3 1,316,000 100.0 ABC
ABC Classification Solution
Stock # Vol. Cost $ Vol. % ABC 206 105 019 144 207 Total 26,000 $ 36 $936,000 71.1 A 200 600 120,000 9.1 A 2,000 55 110,000 8.4 B 20,000 7,000 4 10 80,000 70,000 6.1 B 5.3 C 1,316,000 100.0