Supply Chain Inventory Production Planning Exam Prep Oral

SEMESTER 1- SUPPLY CHAIN INVENTORY & PRODUCTION PLANNING
EXAM QUESTIONS:
SUPPLY CHAIN INVENTORY & PRODUCTION PLANNING FALL 2017
1 FORECASTING. MOVING AVERAGE AND FORECASTING. EXPONENTIAL SMOOTHING FOR A
LEVEL MODEL, A TREND MODEL, AND A SEASONAL MODEL. MEASURES OF FORECAST
ERRORS. (CHAPTER 3)
1.1 FORECASTING
 Forecast combination of an extrapolation of past (statistical forecasting) and informed judgments
about future events (due to internal and external factors)
 required in all parts of company (planning & scheduling, capacity planning, sales planning,
production planning, budget planning…)
 forecasts likely to have errors due to ex post view
Models
 additive model:
useful when the seasonal variation is relatively constant over time
 multiplicative model:
useful when the seasonal variation increases over time
 mixed (additive and multiplicative)
Steps involved
1. Select an appropriate underlying model
some models only for short-/medium-/long-term forecasting; managerial knowledge
2. Select the values for the parameters
sometimes no data available; subjective basis essential
3. Use the model to forecast future demands
1.2 MOVING AVERAGE AND FORECASTING (MEDIUM RANGE)

 Based on Level Model
𝒙𝒕 = 𝒂 + 𝝐 𝒕
 forecasting value derived by using specific amount of historical data (e.g. N=5  last five periods’
average)
1
1.3 EXPONENTIAL SMOOTHING FOR A LEVEL MODEL, A TREND MODEL, AND A SEASONAL MODEL.
a= level
b= (linear) trend
Ft = seasonal coefficient (index) appropriate for period t
ϵt = independent random variables
1.3.1 Exponential for Level Model (smoothed average)

 Based on Level Model
𝒙𝒕 = 𝒂 + 𝝐 𝒕
 minimising discounted sum of squared variables
 intuition is to put less weight on older data

 Uses moving average and start in middle of periods considered for moving average
 the higher α is, the more reliable is current period results for next forecast
 225.7 is forecast for any following period
1.3.2 Exponential for Trend Model

 Based on trend Model
𝒙𝒕 = 𝒂 + 𝒃 ∗ 𝝉 + 𝝐𝒕
 τ (tau) determines to which period to forecast (e.g. if in period 6 and forecast is
period 10 τ)
 minimising least squared regression analysis
, leading to:
2
 these are then updated based on new data
1.3.3 Exponential for Seasonal Model

 Based on Multiplicative trend-seasonal model
𝒙𝒕 = (𝒂 + 𝒃 ∗ 𝝉) ∗ 𝑭𝒕 + 𝝐𝒕
 â and b^ derived with regression as before
 Seasonal factor Ft must first be normalised using moving average and centred average
 Updating through
 Alpha, beta and gamma..

o Small values put little weight to recent dataappropriate in stable condition
o Large values put little weight on historyappropriate in unstable conditions
1.4 MEASURES OF FORECAST ERRORS.

 Forecast’s accuracy: how far away our point estimate forecast is from the actual value
Average Error
Mean Square Error (MSE) (Variance)

 For one-period-ahead forecasts
 When updating the MSE, one should use exponential smoothing
3
 MSEt requires less storage of data (only most recent used)

 gives equal weight on every period in history
Mean Absolute Deviation (MAD)

 For one-period-ahead forecasts
 good statistic to use when analysing the error for a single item
Mean Absolute Percentage Error (MAPE)
 good if one doesn’t know an item’s typical demand volume

 easy to compare
Standard Deviation
0
assuming that MSE is unbiased, sigma is important to set safety stock levels
4
2 ORDER QUANTITIES WHEN DEMAND IS APPROXIMATELY LEVEL. GENERAL ASSUMPTIONS,

DERIVATIONS OF EOQ MODEL, EPQ MODEL, AND EOQ WITH QUANTITY DISCOUNT.
(CHAPTER 4)
2.1 ASSUMPTIONS
1. demand rate is constant and deterministic approximate level known
2. cost factors (e.g. A and r) rather constant
3. Q = integral number of units
4. no discounts (unit variable cost v interdependent from replenishment quantity (no discounts)
5. No shortages
6. Replenishment lead time L = 0
7. item is treated independently of other items
8. Time variable of notations should be the same (convert accordingly)
5
2.2 EOQ MODEL (ECONOMIC ORDER QUANTITY)
6
 sensitivity analysis can give us info how costs will change when deviating from (EO)Q*
 p = (Q/EOQ)-1
 Percentage of Cost Penalty (PCP)
2.3 EPQ MODEL

 finite replenishment economic order quantity (FREOQ), also known as Economic Production
Quantity (EPQ)
7
 Assumption Replenishment quantity (Q) arrives at a rate of m units per time
𝑫
𝑨𝑫 𝑸(𝟏− )𝒗𝟎 𝒓 𝑫
 𝑻𝑹𝑪(𝑸) = + 𝒎
+ 𝑫𝒗𝟎 𝟏 − 𝒎 correction factor
𝑸 𝟐
𝟐𝑨𝑫 𝟏
 𝑬𝑷𝑸 = √ 𝑫 = 𝑬𝑶𝑸 ∗
𝒗𝟎 𝒓(𝟏− ) 𝑫
𝒎 √(𝟏− )
𝒎
2.4 EOQ WITH QUANTITY DISCOUNT

 Different unit costs in all units’ discount
 For the discounts, the Dv0 must be included in the TRC now as it is essential for determining the
replenishment quantity
eventually, ending up with two cost functions:
𝑨𝑫 𝑸𝒗𝟎 𝒓
𝑻𝑹𝑪(𝑸) = + + 𝑫𝒗𝟎  for 0 < Q ≤ Qb
𝑸 𝟐
𝑨𝑫 𝑸𝒗𝟎 (𝟏−𝒅)𝒓
𝑻𝑹𝑪𝒅 (𝑸) = + + 𝑫𝒗𝟎 (𝟏 − 𝒅)  for Qb ≤ Q
𝑸 𝟐
finding the best Q to reduce the costs accordingly

1. Compute the EOQ when the discount d is applicable, that is:
2. Compare EOQ(d) with Qb:

a. EOQ(d) ≥ Qb, then EOQ(d) is the best order quantity
b. EOQ(d) < Qb, then do step 3

3. Evaluate the TRC(EOQ) and TRC(Qb):
a. 𝑻𝑹𝑪(𝑬𝑶𝑸) = √𝟐𝑨𝑫𝒗𝒓 + 𝑫𝒗𝟎 (Dv0 must be included to compare with discount)
𝑨𝑫 𝑸𝒗𝟎 (𝟏−𝒅)𝒓
b. 𝑻𝑹𝑪𝒅 (𝑸𝒃 ) = + + 𝑫𝒗𝟎 (𝟏 − 𝒅)
𝑸 𝟐
compare the two costs accordingly to choose the quantity with smaller costs:
8
If TRC(EOQ) < TRC(Qb), ordering EOQ is best order If TRC(EOQ) > TRC(Qb), ordering Qb is best order
quantity, by using quantity
𝟐𝑨𝑫 𝟐𝑨𝑫
𝑬𝑶𝑸 = √ 𝑬𝑶𝑸𝒃 = √
𝒗𝟎 𝒓 𝒗𝟎 (𝟏 − 𝒅)𝒓
when we have discount, then v0 will be smaller, leading to EOQ1 being bigger as v0 is in the
denominator
9
3 LOT SIZING FOR INDIVIDUAL ITEMS WITH TIME-VARYING DEMAND. GENERAL ASSUMPTIONS.
THE WAGNER-WHITIN METHOD. THE SILVER-MEAL HEURISTIC. OTHER PROCEDURES.
(CHAPTER 5)
demand rate deterministic, but amount varies with time
 we can no longer assume that best strategy is always to use the same replenishment quantity
 now use the demand information over a finite period, which is known as horizon
horizon should be as short as possible
 three approaches to deal with deterministic, time-varying demand pattern:
1. Use of the basic EOQ
based on average demand rate out of horizon; makes sense when variability of demand is low
2. Use of the exact best solution to a particular, mathematical model of the situation
Wagner–Whitin (1958) algorithm; only under specific assumptions
3. Use of an approximate or heuristic method
capturing the essence only but necessarily being the optimal one
3.1 GENERAL ASSUMPTIONS

1. Demand rate D(j) known but varies over time
2. D(j) available at the beginning of period j
3. cost factors (e.g. A and r) rather constant
4. Q = integral number of units
5. no discounts (unit variable cost v interdependent from replenishment quantity (no discounts)
6. No shortages
7. Replenishment lead time L = 0 (or known and deterministic)
9. Carrying cost to inventory only when carried over from one period to next
 only 1,2 and 9 are different to EOQ-Model
10. Time variable of notations should be the same (convert accordingly)
3.2 THE WAGNER-WHITIN-METHOD

 Wagner and Whitin (1958) developed an algorithm that guarantees an optimal solution
 two additional assumptions necessary for optimal solution
Property 1: repl. when inventory is zero
Property 2: upper limit to how far of period j we include repl. needed
 Can easily be set up in Excel but otherwise generally quite time consuming
10
Drawbacks:
1. relatively complex nature (difficult for practitioner to understand than other approaches)
2. need for a well-defined ending point for the demand pattern
3. MRP software operates on a rolling schedule Q chosen should not change when new
information about future demands becomes available
Wagner–Whitin does not have this property
4. assumption that replenishments can be made only at discrete intervals (namely, at the beginning of
each of the periods)
3.3 THE SILVER-MEAL HEURISTIC

 determine time UC to find periods when to order how much, based on local minimum of the
TRCUT (Total Relevant Cost Unit Time) T-th period
 order costs + carrying 2nd periods demand one period
 order costs + carrying 2nd periods demand one period AND demand of 3rd period for two periods
 eventually, finding FIRST LOWEST UNIT COST per time before it rises again and replenish amount
need until this period (in graphic e.g until 3)
 afterwards start procedure from next period again (in graphic e.g. 4)
11
Drawback
 problem (compared to Wagner-Whitin): Silver-Meal only looks at short period but problem must be
considered over whole periods available
only local minimum (3) is considered for decision, but not global minimum (5)
3.4 OTHER PROCEDURES

3.4.1 L4L (Lot-for-Lot)
 simply orders exact amount needed for each time period
Reasonable when inventory carrying cost is high and set-up cost is low
3.4.2 EOQ Model

 If demand rate is approximately constant (little variation), fixed EOQ should perform suitable
𝟐𝑨𝑫
𝑬𝑶𝑸 = √
𝒗𝟎 𝒓
D̅: average demand rate out of horizon N
 EOQ should be adjusted to exactly satisfy requirements (nearest cumulative demand)
3.4.3 Least Unit Cost (LUC)

 Procedure like Silver–Meal, but here, total UC are used as criterium (rather than cost per T period)
3.4.4 Part-Period Balancing (PPB)

 select number of periods covered by the replenishment such that total carrying costs are made as
close as possible to setup cost
 rather order 60 units than170

 PPB heuristic generally performs poorly in dynamic environments
3.4.5 Periodic Order Quantity Supply (POQ) EOQ Expressed as a Time Supply
 EOQ expressed in terms of time T
12
result rounded to nearest higher integer

For Example:
 at EOQ order size, the quantity lasts to cover demand for 1.94 periods
 round to 2 periods  order must cover two period
3.5 WHEN TO USE HEURISTICS

 EOQ is best used when demand is not variable
 variability should exceed some threshold before heuristic is used
to determine variability, one can use squared coefficient of variation (SCV)
 tests show, if SCV<0.2, then use EOQ and if SCV>0.2, use heuristic
13
4 INVENTORY MANAGEMENT FOR CLASS-B ITEMS WITH PROBABILISTIC DEMAND. THE (S,Q)-
SYSTEM, THE (S, S)SYSTEM, THE (R, S, S)-SYSTEM, AND THE (R, S)-SYSTEM. COST AND
SERVICE OBJECTIVES. HOW TO CHOOSE S IN (S,Q)-SYSTEM WITH A GIVEN Q. (CHAPTER 6)
Key issues to solve:

1. How often inventory status should be determined? (continuous vs periodic)
2. When replenishment order should be placed? (s?)
3. How large replenishment order should be? (Q)
 Manager must consider following questions to answer the previous three:

1. How important is item (A, B, C)?
2. Stock status be reviewed continuously or periodically?
3. What form should inventory policy take [(s, Q), (s, S), (R, s), (R, s, S)]?
4. What specific cost or service objectives should be set?
Definitions:
1. On-Hand (OH) stock
 stock physically on shelf (never negative)
2. Net Stock
3. Inventory Position/ Available Stock
on-order: already ordered but not received yet

committed: cannot be used in short-run as already planned in usage
4. Safety Stock (SS)
defined as average level of net stock just before a replenishment arrives
What, if demand can not be satisfied? Two extreme cases:

1. Complete backordering
demand is backordered and filled as soon as an adequate-sized replenishment arrives
2. Complete lost sales
demand is lost as customer will go to competitors (e.g. customer will not backorder a bread)
Continuous Periodic
Advantage Disadvantage Advantage Disadvantage
 Satisfy demands  More expensive, e.g.  Less  Higher SS
at any time setting the correct expensive  Possibly running out of stock
 Less SS variables during the non-review-time
4.1 THE (S,Q)-SYSTEM

 Review cycle (R)=0continuous review system
14
 Fixed Q is ordered when inventory drops to reorder point s or lower

inventory position and NOT net stock is used to trigger order
Advantage Disadvantage
 Simple, esp. in two-bin-form  In unmodified form it may not effectively
 if material for production in one cope with situation of large transactions
depleted, order is release; 2nd bin has esp. when Q will not rise inventory
enough material to cover demand level above s-point, Q may be too less
4.2 THE (S, S)-SYSTEM

 R=0continuous review system
 If demand transactions is one unit, the (s, S) = (s, Q)
 variable Q is ordered when inventory drops to reorder point s or lower
raise inventory level above s
 inventory always between min (s) and  Variable Q
max (S) potentially more errors by suppliers
who prefer fixed order quantity (also
regarding packaging or handling)
4.3 THE (R, S)-SYSTEM

 Aka replenishment cycle system
 Every R units of time, enough is ordered to fil up to fill up inventory level to S
 Savings due to less reviews  Variable Q
 Adjustable “order-up-to-level S” at potentially more errors by suppliers
every R  Higher carrying costs than in
good, if demand changes over time continuous-review systems
4.4 THE (R, S, S)-SYSTEM

 a combination of (s, S) and (R, S) systems
o (s S) special case where R=0
o (R, S) special case where s = S-1
 idea: every R units of time check the inventory position and evaluate:
a) at or below reorder point s?  then order enough to raise it to S
b) above s?  nothing is done until at least next review
 using some assumptions Scarf (1960)  high computational effort to obtain best
showed that best “R, s, S”-system values of three parameters (R, s, S)
produces a lower total of
replenishment, carrying, and shortage
costs than does any other system
4.5 COST AND SERVICE OBJECTIVES

1. SSs based on Simple-Minded Approach
 assigning common safety factor
2. SSs based on Minimizing Cost
cross checking with higher transport cost vs extra inventory carry cost
15
3. SSs based on Customer Service

 fulfil certain service goal (rather have inventory cost than shortage cost)
4. SSs based on Aggregate Considerations
 provide the best possible aggregate service across a population of items
4.5.1 Simple-Minded Approach

1. Equal Time Supplies
 Order are released once inventory position minus forecasted lead time drops below, e.g. 2-months-
supply
2. Equal Safety Factors
 SS as product of two factors:
𝑺𝑺 = 𝒌 ∗ 𝝈𝑳
(mean assumed to be 0)
k: safety factor
σ: std. deviation of the errors of forecasts of total demand
4.5.2 Minimizing Costs

1. Specific costs (B1) per stock out occasion (fixed amount)
2. Specific, fractional charge (B2) per unit short
costs per unit short of an item i is: B2vi units short made through e.g. overtime production
3. Specific, fractional charge (B3) per unit short per unit time
B3 per DKK short (or B3v)
4. Specific charge (B4) per customer item line short (fixed amount as per Service-Level-
Agreement)
4.5.3 Customer Service

1. Specified probability (P1) of no stock out per replenishment cycle – Cycle Service Level
2. Specified fraction (P2) of demand to be satisfied routinely from the shelf – Fill Rate
Decision rule for B3 shortage costing measure is equivalent to that for the P2 measure with:
𝐵3
𝑃2 =
𝐵3 + 𝑟
3. Specified fraction (P3) of time during which net stock is positive – Ready Rate
4. Specified average time (TBS) between stock out occasions
4.5.4 Aggregate Considerations

1. Allocation of a given total safety stock (TSS) among several items to minimize expected total
stock-out occasions per year (ETSOPY)
Rule is identical to that obtained by assuming a value of B1 (the same for all items), and then
selecting the safety factor (k) to keep the total of carrying and stock out costs as low as possible
2. Allocation of a given total safety stock (TSS) among several items to minimize expected total
value of shortages per year (ETVSPY)
Leads to a decision rule for selecting the safety factor of each item based on B 2
4.6 HOW TO CHOOSE S IN (S,Q)-SYSTEM WITH A GIVEN Q

4.6.1 Assumptions
1. Demand probabilistic but with constant average demand rate
2. Replenishment orders placed when inventory position equals s
 assumes demand of unit size or that undershoots are only of marginal importance
3. Orders arrive in same time sequence as ordered
4. unit shortage costs = high
 average level of backorders will be negligible compared to average level of stock on hand
5. Forecast errors have a normal distribution with no bias and a known standard deviation σL over
replenishment period
16
6. (Optimal) order size Q is predetermined (we focus on B-class items)

7. Cost of the control system does not depend on s
4.6.2 General approach of finding s when Q given (then used B1 cost objective)
(relation of s to SS)
 The forecast provides estimates of:

a) Demand during the lead time L, i.e. x̂L
b) Standard deviation of demand during the lead time period L, i.e. σL
using the above formulas, determination of an optimal k value leads directly to a value of optimal
s (computing s via k)
(Management must specify this value because inventory calculations may recommend, in effect,
holding no inventory at all)
optimal s defines optimal policy as (EO)Q is predetermined by at this point
 Basic components of the Expected Total Relevant Cost (

ETRC(k) = (Replenishment Cost) + (Holding Cost) + (Stock out Cost)
𝑬𝑻𝑹𝑪(𝒌) = 𝑪𝒓 + 𝑪𝒄 + 𝑪𝒔
 optimal k can be found by standard differentiation of the ETRC(k) and setting the derivative equal to 0
 ETRC in specific have common Cr and Cc but differ in Cs depending on cost/ service
objective!!!
4.6.2.1 Replenishment Costs

 Expected demand divided by (optimal) Q gives amount of replenishments per year
  multiply with A
𝑨𝑫
𝑪𝒓 = ∗ 𝑫𝒗
𝑸
4.6.2.2 Inventory Costs
 From basic definition
Net Stock= On Hand – Backorders
follows
E(NS)= E(OH) -E(BO)
17
 Using assumption 4 having no backorders and mean rate of demand being constant with time:
on average, the OH-level drops from arrival of new replenishment (s-x̂L+Q) to SS (s-x̂L) right
before next replenishment arrives
 expected inventory
 leading to cost…
𝑸
𝑪𝒄 = ( + 𝒌𝝈𝑳 ) ∗ 𝒗𝒓
𝟐
4.6.2.3 Stock-out Costs
 For cost objective B1 (cost per stock-out occasion):
𝑫𝑩𝟏
𝑪𝒔 = ∗ 𝒑𝒖≥ (𝒌)
𝑸
1. Expected # of replenishments per cycle (i.e. D/Q)
2. Cost per stock out (B1)
3. Probability of stock out per cycle (i.e. pu≥(k))
when calculating, use […]*1 -Norm.S.Vert(k)
 Compared to e.g. cost objective B2 (cost per unit short).

𝑫𝑩𝟐 𝒗𝝈𝑳 𝑮𝒖 (𝒌)
𝑪𝒔 =
𝑸
1. Expected # of replenishments per cycle (i.e. D/Q)
2. Expected shortage per replenishment cycle σLGu(k)
3. Cost per unit short B2v
4.6.2.4 Finding optimal k (for B1)

 how to find the optimal k now without using solution finder in spreadsheet?
Step 1: Evaluate, if
𝐷𝐵1
<1
√2𝜋 ∗ 𝑄𝑣𝑟𝜎𝐿
 Q predetermined in chapter 4; B1 is expressed in currency; all other variables defined in 6.7.1
if left side<1, go to step 2, otherwise solve for k
𝐷𝐵1
𝑘 = √2 ln ( )
√2𝜋 ∗ 𝑄𝑣𝑟𝜎𝐿
 if k is lower than minimum specified by management, then go to step 2, otherwise go to step 3
Step 2: set k equal to lowest value allowed by management
Step 3: set reorder point equal to s = x̂L + kσL, rounded to nearest integer (if step 2 used, round to highest
integer)
4.6.2.5 Finding optimla k (for B2)

Step 1: Evaluate, if
18
𝑄𝑟
>1
𝐷𝐵2
 Q predetermined in chapter 4; B2 is expressed in currency; all other variables defined in 6.7.1
if left side>1, go to step 2, otherwise select k so to satisfy
𝑄𝑟
𝑝𝑢≥ (𝑘) =
𝐷𝐵2
Using in Excel NORMSINV(1 – pu≥(k))
 if k is lower than minimum specified by management, then go to step 2, otherwise go to step 3

Step 2: set k equal to lowest value allowed by management
Step 3: set reorder point equal to s = x̂L + kσL, rounded to nearest integer (if step 2 used, round to highest
integer)
4.6.2.6 Additional Info Derivation of k
19
4.6.3 Common Derivation

 When demand during L is normally distributed has PDF: fx(x0):
Figure 1: Normally distributed forecast errors
 inventory position of s units at time t must have reached the stocking shelf by time t +L
 Demand during lead time has PDF: fx(x0)
three important results emerge:
Safety Stock (SS): expected net stock (ENS) just before the next replenishment arrives
Prob. Stock out during L: probability that demand in lead time is at least as large as s (see figure 1)
expected shortage per replenishment cycle (ESPRC): x0-s is the stock-out (OH=0)
20
 when forecast errors are assumed to be normally distributed and the SS is expressed as SS=k σL, then
1. and 3. Become:
21
5 INVENTORY MANAGEMENT FOR CLASS-A ITEMS. DETERMINATION OF PARAMETERS OF (S,Q)

SYSTEM; (S, S) SYSTEM WITH UNDERSHOOTS. (CHAPTER 7)
 A-item most important items low volume but high in value

importance determined on high annual dollar usage (D*v)
Dv important factor whether to put item in A category or not
 costs of replenishment, carrying stock, and shortages associated with such an item are high enough
to justify a more sophisticated control system
In essence, trade-off between control system costs
 Two subclasses
o Low v, high D (chapter 7)
o High v, low D (chapter 8)
 A items to be observed continuously storage cost might be extreme
 Estimation and influencing the demand actively
o E.g. manual inputs (calling customers)
o Anticipate demand (based on historical data)
o Manipulate demand
Recall CHP6
𝒔 = 𝒙𝑳 + 𝒌𝝈𝑳
𝑨𝑫 𝑸 𝑫𝑩𝟏
𝑬𝑻𝑹𝑪 = 𝑪𝒓 + 𝑪𝒄 + 𝑪𝒔 = ∗ 𝑫𝒗 + ( + 𝒌𝝈𝑳 ) ∗ 𝒗𝒓 + ∗ 𝒑𝒖≥ (𝒌)
𝑸 𝟐 𝑸
Dv neglected because only constant now
 Assumed that demand over lead time is normally distributed
5.1 DETERMINATION OF PARAMETERS OF (S,Q) SYSTEM

 (s, Q) is a continuous review system that orders Q once the inventory position drops to reorder point
s or below
 For very important items, it may be valuable to simultaneously determine s and Q
 how determine s and Q simultaneously?!
 by solving cost minimization problem of “min ETRC (s,Q)”
note that s corresponds to safety factor k, hence we solve for
 To reduce number of separate parameters that must be considered, both sides normalized by
multiplying by constant term 2/vrσL
 result as: normalized total relevant costs, NTRC(k, Q)
As per Appendix Chapter 7, this will lead to the following equations:
22
Iterative procedure:
1. Find EOQ initially and set Q=EOQ for “Equation k”
𝟐𝑨𝑫
𝑬𝑶𝑸 = √
𝒗𝟎 𝒓
23
2. Find the value for k for iteration #1

3. use k from step 2. to compute new Q (“Equation Q”) and k (“Equation k”) for the other iterations
until difference between Q and k is infinitesimal
 The first iteration number was sequential (first Q=EOQ and then k was determined); thereafter
simultaneously k and
5.2 DETERMINATION OF PARAMETERS OF (S, S) SYSTEM WITH UNDERSHOOTS

5.2.1 Simple Sequential Determination of s and S
 Characterization:
o continuous inventory review
o when inventory drops to/below s, order is placed to raise inventory to S
o (additional assumption: transactions are in unit size)
this leads to order happening strictly at s, and therefore Q is:
Q = S-s
this lead evaluates in (s,S) being like (s,Q)
 it is undershot that complicates the analysis
5.2.2 Simultaneous Selection of s and S Using the Undershoot Distribution

 Undershoot  when the replenishment order is submitted, inventory level is less than s
 z = amount of undershoot
 x’= z + x (total amount needed during lead item)
x’ = normal random lead period demand with mean x̂L and standard deviation σL
1. we know the probability distribution of the transaction size
2. when S − s is considerably larger than average transaction size, we can make use of a result
developed by Karlin (1958):
undershoot distribution can be approximated as follows:
pz(z0) = probability that the undershoot is of size z0

pt(t0) = probability that a demand transaction is of size t0
E(t) = average transaction size
 assumptions 1. + 2. Lead to mean and variance:
24
3. assuming that z and x are independent, we derive:
 x’ assumed to have normal distribution with mean and variance given above
Decision rules procedure:
Step 0: compute the expectation E(x’) and variance Var(x’)
Step 1: Select k and Q to simultaneously satisfy the following two equations
(rounded to an integer) and
with
2𝐴𝐷
𝐸𝑂𝑄 = √ and 𝜎𝑥′ = √𝑣𝑎𝑟(𝑥 ′ )
𝑣𝑟
Step 2: 𝑠 = 𝐸(𝑥 ′ ) + 𝑘𝜎𝑥′ rounded to nearest integer
Step 3: 𝑆 = 𝑠 + 𝑄
(Step 4: calculate the ETRC)
 ETRC of simultaneous with undershoot is very close to optimal solution found by Archibald, 1976
25
6 THE STYLE GOODS PROBLEM. THE UNCONSTRAINED, SINGLE-ITEM, NEWS VENDOR PROBLEM.
DERIVATION OF DECISION RULE BY MARGINAL ANALYSIS. A DERIVATION OF THE SAME RESULT
BASED ON PROFIT MAXIMIZATION – AN INTUITIVE DERIVATION IS PERFECTLY LEGITIMATE. THE
SINGLE PERIOD, CONSTRAINED MULTI-ITEM SITUATION. POSTPONED PRODUCT
DIFFERENTIATION. (CHAPTER 9)
 Previous chapters concerned with models where goods were stored in inventory from one selling
period to another
Possibility to store removed because of perishability or obsolescence
 main features of style goods problem:
o Short selling season
o Orders must be put prior to the selling season
o Uncertain forecast (last data maybe from previous season and not up-to-date)
o Costs associated with stockouts (urgency deliveries, lost revenue, reputation)
o Overshoot cost prevail when too much is ordered (items have lower value after season)
6.1 THE UNCONSTRAINED, SINGLE-ITEM, NEWS VENDOR PROBLEM

 A newsvendor must determine Q of copies of a particular newspaper to purchase (to stock)
As before, problem is:
Q newspapers VS
x of random demand subject to
o Underage cost cu (of demand that cannot be met)
o Overage cost co (for each copy not being sold)
o px<(x0) (probability that total demand x is less than x0)
6.2 DERIVATION OF DECISION RULE BY MARGINAL ANALYSIS
 increasing the Q will result in lower underage probability and cost (cu) but higher overage probability
and cost (co)
finding the Q* where cu=co!!!
 expected overshoot cost associated with acquiring Qth unit should be equal to the expected
undershoot cost saved
26
6.3 A DERIVATION OF THE SAME RESULT BASED ON PROFIT MAXIMIZATION – AN INTUITIVE

DERIVATION IS PERFECTLY LEGITIMATE
Under Perfect information (knowing the demand)
v = acquisition cost per unit
p = revenue per unit sold
B = penalty per unit not satisfied demand
g = salvage value per unit not sold
 Inserting the two profit functions into Expected Profit leads to
 Using
leads to (under perfect information!!!)
 Finding optimal Q by taking the derivative of E[P(Q)], we find
27
Under normal distributed demand
pu≥(k) is the probability that a unit normal variable takes on a value of k or larger
 Expected profit in normally distributed demand transforms into:
6.4 THE SINGLE PERIOD, CONSTRAINED MULTI-ITEM SITUATION

 Assumptions:
o More than one SKU to be stocked for a single period’s demand
o Demand for each SKU independent of the others
o budget or space constraint on the group of items
different newspapers share a limited budget - budget constraint
amount of repair kit taken by the repair crew - space constrain
 Maximize expected profits with a given budget for acquisition
28
this optimization problem can be solved using the Lagrange-Method:
Any given M yields a corresponding order Qi for each item i = 1, , n, which in turn allows to evaluate
the corresponding budget usage
Resulting decision procedure is:

Step 1: Select an initial positive value of the multiplier, M
29
Step 2 Determine each Qi (i = 1, 2, 3, ..., n) to satisfy
Step 3
Excel’s “Goal Seek” is helpful to find optimal M that satisfies constraint
 in case a quantity constraint of W units is present, then the Lagrangian function changes in Step 2:
Step 1: Select an initial positive value of the multiplier, M
Step 2 Determine each Qi (i = 1, 2, 3, ..., n) to satisfy
Step 3: as above, we compare
as above and act, depending on the results
6.5 POSTPONED PRODUCT DIFFERENTIATION

6.5.1 One final product made from base product
 Producing basic product and differentiate later as per demand
approach is known as delayed (or postponed) product differentiation or postponement
v0 = cost to produce the base product

t = cost to differentiate the base product
 total cost per product is TCv = v0 + t
g = salvage price for selling finished product
g0 = salvage price for base product
Costs of underage and overage, amended by the t:
Profit Function Single Product:
30
6.5.2 Multiple end products made from basic product

 Cost of OVERSTOCKING remain the same (firm only assumes to have shortages), then it
again follows as before
 Cost for UNDERSTOCKING will differ among items (each has different customisation
with cost)
 ESi for each item leads to new profit function
 To avoid solving multivariable optimisation problem, the following simplification is assumed:

1. differentiation takes place upon realisation of corresponding demand
 then we may consider the demand-weighted average underestimation cost:
(e.g. if item 1 has twice average demand as item 2, we would give twice weight to cu for item 1)
2. Prioritisation of Demand:
a. we can observe demand first and then choose what type of demand to satisfy
 Naturally, first satisfy demand that corresponds to highest cu and then move
to next highest costs
b. all unsatisfied demand then responds to item with cost
leading to new estimated profit maximisation
 cu remains but co are smaller now as differentiation only occurs for real demand
 overstocking is less severe and thus is willing to purchase more
 Optimal quantity of such special items to order depends on analysis of marginal costs
31
7 COORDINATION OF REPLENISHMENTS FOR MULTIPLE ITEMS AT A SINGLE STOCKING POINT.

REPLENISHMENT QUANTITIES IN A FAMILY OF ITEMS. INCORPORATION OF GROUP DISCOUNTS
FOR THE DETERMINISTIC CASE. THE (S, C, S) SYSTEM. THE SIMPLE ECONOMIC LOT-
SCHEDULING PROBLEM. (CHAPTER 10)
 It may be advantageous coordinate replenishment of items if all are stocked at the same location
and they share a common supplier or mode of transportation or production facility
 Savings on unit purchase costs  possible increase in average inventory level
 Savings on unit transportation costs  increase in system control costs
 Savings on ordering costs  Reduced flexibility
 Ease of scheduling
7.1 REPLENISHMENT QUANTITIES IN A FAMILY OF ITEMS

7.1.1 Assumptions
1. demand rate is constant and deterministic (7. + 8.)
2. Q does not need to be integral number
3. no discounts (v does not depend on Q)
4. cost factors do not change significantly with time (e.g. A, r)
6. No shortages
7. L = 0 (Replenishment lead time=0)
8. Inventory replenished when 0
9. Entire order delivered at the same time
(mi = integer number of T intervals that replenishment quantity of item i will last)
A= cost of placing order for items in family (independent of the number of distinct items involved)
ai = minor setup cost associated with including item i in a replenishment (also called line cost; cost for
adding one more item in the line)
Recall:
where TEOQ translates into
32
 major A occurs every T and minor a occurs every mT:
TRC depend on T and m simultaneously!!

 first find T (together with D, the m can be found)
Finding T*
 substituting this into TRC leads to
Finding m*
 Derivation of TRC*(mi) is complicated nonlinear problem as ms influence each other and
leads to:
which means:
Assumptions
1. that the same relation holds even when m i are restricted to integers
𝑎
2. item 1 in our notations corresponds to the smallest ratio 1 and let m1 =1
𝐷1 𝑣1
 then after further algebra, we find
Decision Procedure
𝑎1
Step 1 Number the items such that is smallest for item 1. Set m1 = 1
𝐷1 𝑣1
33
Step 2 Evaluate the other mi’s
rounded to the nearest integer greater than zero

Step 3 Evaluate T* using the mi’s of Step 2 in Equation
Step 4 Determine
Cost interpretation
 cost of using TRC (T,mi) cannot be optimal due to rounding of values
 however, a bound is given by
o first term is TRC EOQ strategy for item 1 alone

o second term is summation of TRC EOQ strategy of each of other items
7.2 INCORPORATION OF GROUP DISCOUNTS FOR THE DETERMINISTIC CASE

 we know from preceding that an m1 > 1 is likely for item having a high value of ai/Divi;
such items should not necessarily be included in every replenishment, even to help achieve a
quantity discount level
 Step 1
𝑎1
Step 1.1 Number the items such that is smallest for item 1. Set m1 = 1 and compute the
𝐷1 𝑣1
mi’s and T as before, but assuming that
v0i = basic unit cost of item i without a discount

d= fractional discount when the total replenishment equals or exceeds the breakpoint quantity Qb
 Step 1.2
(it does not matter if discount is included in computing the m’s as it will cancel out)
 Step 1.3 Qsm in currency, thus

Qi*v0i!!!
 If Qsm ≥ Qb, use the mi’s, T and Qi’s found above. If not, proceed to Step 2.
 Step 2 Scale up the family cycle time T (found in step 1) until the smallest replenishment size equals
the quantity breakpoint:
34
The mi’s of step 1 are maintained to compute the cost of this breakpoint
 Step 3 Use the procedure of Section 10.2 to find the ii’s, T and Qi’s without a discount and compute
the TRC, using
OR
 Step 4 Compare the TRC values found in Steps 2 and 3 and use the m’s, T and Q i’s associated with
the lower cost
7.3 THE (S, C, S) SYSTEM.

 Questions to be answered include:
o How often do we review the status of items (choice of R)?
o When do we reorder the group?
o How much do we order?
Main problem: how estimate average inventory levels and service implications of individual items
 whenever item i’s inventory position drops to or below si (must-order point), it triggers
replenishment that raises item Si
 at same time, any other item j (within the associated family) with its inventory position at or below its
can-order point cj is included in the replenishment with Q that raises inventor to Sj
7.4 THE SIMPLE ECONOMIC LOT-SCHEDULING PROBLEM

 How should the factory schedule the production to meet replenishment goals?
 Economic Lot Scheduling Problem (ELSP)
 approach of production of several items is to create a sequence of production that will be repeated
periodically
 ELSP problem finds:
o cycle length
35
o production sequence
o production times
o idle times
such that such that demand during cycle is satisfied, inventory and setup costs are minimized
Major assumption: the demand is deterministic
 two constraints/ difficulties:
1. satisfy a production capacity constraint
2. have only one product in production at a time (“synchronization” constraint)
 setup plus processing times shall be smaller than cycle T (DiT is the batch size for item i)
36
8 MULTIECHELON INVENTORIES. SEQUENTIAL STOCKING POINTS, POLICIES AND ECHELON

STOCK. COMPLICATING FACTORS IN THE CASE OF PROBABILISTIC DEMAND, BASIC MODELS
AND MAIN CONCLUSIONS. (CHAPTER 11)
 In multiechelon inventory systems retail outlets are supplied from branch warehouses, which in
turn are supplied from central warehouses
Figure 2: Three-echelon illustration

Retail outlets (the first echelon) are replenished from branch warehouses (the second echelon)
 framework for supply chain management includes structural and coordination decisions
o Structural decisions
on long-term, deterministic approximations
 Where locate facilities and how many?
 What shall be size of facility?
 Which facility should produce what and how much?
 Modes of transportation for products?
Result: a network of facilities designed to produce and distribute products
o Coordination decisions (taken after structural decision are made)

take structure of the multiechelon network as given and focus on the short term
 Inventory stocking and replenishment decisions be made centrally or in a
decentralized
 Can, or should, the stock status be reviewed continuously or periodically?
 What form should the inventory policy take [(s, Q), (s, S), (R, s), (R, s, S)]?
 What specific cost or service objectives should be set?
8.1 SEQUENTIAL STOCKING POINTS, POLICIES AND ECHELON STOCK.

 Eternal deterministic demand given
 stocking points are serially connected, e.g. one central warehouse, one retailer warehouse, and
one retail outlet
37
Assumptions:
 Lead = 0 at both levels
 Nested type of replenishments:
 replenishment at warehouse immediately triggers replenishment at retailer
 inventory of warehouse (IW ) does not undergoes sawtooth pattern when retailer withdraws Q R
 instead of using average inventory (complicated to compute), artificial echelon stock used
 Echelon Stock of echelon j (in a general multiechelon system) is the number of units in the system
that are at, or have passed through, echelon j but have as yet not been committed to outside
customers
 specific echelon inventory valued at value added
 in two stage echelon, value of echelon inventory is as follows:
o Warehouse:
o Retailer:
I̅’W = average value of the warehouse echelon inventory, in units

I̅’R = average value of the retailer echelon inventory, in units
 two variables QR and n to be optimised
38
Decision Rule:
Step 1: Compute
If n*is exactly an integer, go to Step 4 with n = n*

Also, if n* < 1, go to Step 4 with n = 1
Otherwise, proceed to Step 2
Step 2: Ascertain the two integer values, n1 and n2, that surround n*; and go to step 3
Step 3: Evaluate
Step 4: Evaluate
Step 5: Compute
39
8.2 COMPLICATING FACTORS IN THE CASE OF PROBABILISTIC DEMAND, BASIC MODELS AND MAIN
CONCLUSIONS
 system remains essentially the same, with lead times explicitly introduced
 lead times condition on upper level having an adequate stock!
 Straightforward use of the methods of CHP 4 through 6 independently make replenishment

decisions based on its own
1. Relevant costs and service considerations
2. Predicted demand (based on historical data)
3. Replenishment lead time from the next stocking point upstream
various flaws come into focus:
1. Lead time is as announced only if there is stock at upper level
if not, lead time of upper level added
2. While demand at lower level can be smooth, demand at higher levels is less frequent and discrete
Questions:
1. How define service and how safety stocks on upper echelons affect stockouts at lower ones?
2. If stock upstream < Order Size: wait for replenishment at upper level or make partial replenishment?
3. Possibility of emergency replenishments between nonadjacent echelons
4. Possibility of replenishments within the echelon
 Two dimensions can be considered

o local VS global information
o centralized (push system) VS decentralized control (pull system)
 best solutions are obtained by using global information and centralized control
8.2.1 Base Stock Control System

 Is response to difficulties of each echelon deciding when to reorder based only on demand from the
next lower echelon
 key change: make end-item demand information available for decision making at all stocking
points
requires the use of an effective communication
 common type of base stock system order-up-to (S) policy (e.g. (s,S))
 base stock level computed as
40
8.2.2 Serial Situation

 computing Q and s that explicitly takes account of the effects of the stock at one echelon on the lead
time at the next lower echelon (De Bodt and Graves (1985))
using global information and centralizes control
two stages echelon (warehouse, retailer)
Prerequisites:
1. demand occurs only at the retailer
normally distributed with x̂ and σ̂ annual parameters
2. deterministic replenishment lead time associated with each stage (LW and LR)
 time LR only begins when sufficient warehouse stock available
3. policy is (s, Q) for either of levels, so requires determining (sW, QW) and (sR, QR)
4. vW and vR are inventory values at different levels
5. B2 penalty appears for value unit short
6. QW = nQR predetermined by deterministic procedure 11.3
7. Reorder point s determined individually as in CH 7 using end-item forecasts for expected lead
time demand appropriate to the echelon under consideration
8. S is determined individually as S = s +Q
9. When echelon inventory ≤ s order up-to-level S is made from the preceding echelon
8.2.3 Arborescent Situation

 How to distribute stock that will not be able to satisfy all retailers?
41
 possibilities of warehouse:
1. No stock at warehouse (break bulk facility) always ship
2. Stock at the warehouse
8.2.3.1 Arborescent not Stock at Warehouse

 Entire amount immediately shipped to retailers; questions:
1. How allocate limited inventory to multiple retailers
2. How much to order each review period
 Centralized control and global information is required
 Decision rule for S and (various) allocation policies are available
 Demand at retailers is independent of each other
8.2.4 Arborescent with Stock at Warehouse

 Even more complicated problem; questions:
1. Amount the warehouse should order from its supplier
2. Amount to ship from the warehouse to the retailers each period
3. Amount to allocate to each retailer when stock runs short
 Centralized control and global information is required
42
9 COORDINATION IN SUPPLY CHAINS, MOTIVATION AND A NEWSVENDOR EXAMPLE. (CHAPTER

12)
The conclusion was: coordination is important
 two main areas where one can find opportunities to improve coordination
o information sharing
o incentive alignment
 poor coordination can occur externally or internally
 One consequence of poor coordination across partner firms in a supply chain is a phenomenon
known as the “bullwhip effect”
 relatively small variability in end-consumer demand can translate to very high variability at
upstream stages in the supply chain
 Possible reasons despite relatively constant demand
o Trade promotions
o Volume discounts
o Long lead times
o Full truckload discounts
o Demand signal processing (if demand ↑ further increases expectations)
o Order batching
 Solutions to improve coordination:
1. Sales and Operations Planning (S&OP)
 help internal information sharing and alignment
2. Collaborative Planning, Forecasting, and Replenishment (CPFR)
 process for external coordination across separate firms
3. Vendor-Managed Inventory (VMI)
 information shared across organizational boundaries & decision-making rights transferred
9.1.1 Sales and Operations Planning (S&OP)

 Effective information sharing is critical to improving supply chain coordination
 key idea behind S&OP
decision makers in a firm should have access to most accurate and up-to-date information
9.1.2 Collaborative Forecasting

 improve information sharing between trading partners
main driver was poor coordination around retail promotional events
promotions occur on days 10 and 25

upper chart
supplier not informed by retailer and simply forecasts
based on historical data (exponential smoothing or
similar)
lower chart
supplier informed by retailer so forecast is adjusted
43
9.1.3 Vendor-Managed Inventory (VMI)

 one cause of bullwhip effect is lack of demand information available to upstream firms
 in VMI: vendor assumes responsibility for determining replenishment quantities for their customers
Key idea:
VMI that permits mitigation of the bullwhip effect
9.2 MOTIVATION OF COORDINATION THROUGH INCENTIVE ALIGNMENT

 How can organize system to make parties’ incentives aligned?
 newsvendor model for illustration. Model consisted of:
o A supplier who sets the newspaper value
o retailer who takes the value and makes the decision about the quantity of newspapers to
order
o Customers with estimated demand for newspapers having mean x^ and st. deviation σ
Retailer’s view
based in which retailer determines optimal Q
 v is only parameter that supplier can coordinate and can agree on with retailer
With increasing v, the ration becomes bigger

 retailer’s ideal order quantity decreases as
wholesale price increases
Supplier’s view
as a function of the wholesale price v and resulting Q*(v) chosen by retailer
9.2.1 Wholesale Price Contract

System’s view
Properties of System
 Incentives not aligned:
o retailer prefers a lower v
o supplier prefers a high v
 supplier’s profit is a deterministic function that does not depend on demand whereas
 retailer bears all uncertainty but only gets some of the profit
 solution: getting supplier to take some of the losses associated with unsold inventory
44
 system does not achieve best overall profit in this arrangement due to double marginalization and
retailer’s risk of excess of inventory
 Total profit function of retailer plus supplier
 wholesale price v drops out and is replace by unit cost of supplier!!!

Q* that maximises total profit does not depend on wholesale price v
<
9.2.2 Buyback Contract

 Lowering retailer’s risk by supplier absorbing some
 supplier can offer selling price vb with buyback option (supplier buys back unsold items of retailer)
 only if b > g (if not, retailer would prefer salvage option)
Observation:
 Clearly, when b increases, retailer prefers to buy more as profit increases
 Surprisingly, the supplier prefers the same
as buyback offer increases, supplier does take on more of the overstocking risk
this shift away from retailer encourages retailer to buy more (and thus increasing total profit)
45
 (v replaced by vb and g replaced by b)
Supplier’s Profit
 excess inventory not necessarily has to return physically to supplier but rather compensate
difference
 compensation to the retailer is sometimes known as “markdown money”
9.2.3 Revenue-Sharing Contract

 Another way to help distribute risk better
 retailer agrees to share some fraction (1−s) of revenue from products sold in exchange for lower
wholesale price vs
 This leads to following cu and co, as well es Q and profits
9.2.4 Challenges of Implementing Coordinating Agreements

Challenges include e.g:
1. Models depend on parties’ make optimal decisions (not always the case)
2. Information symmetry e.g. about cost, process, demand etc.
3. Agreement transactions depend on information that should be transparent, available, and observable
by two parties
4. Multiple products, customers, locations
46
10 THE MEDIUM RANGE AGGREGATE PRODUCTION PLANNING PROBLEM. THE COST

COMPONENTS. LEVEL AND CHASE STRATEGIES. AN LP FORMULATION OF THE PROBLEM.
COMPARISON OF THE METHODS; ADVANTAGES AND DISADVANTAGES. (CHAPTER 14)
10.1 THE MEDIUM RANGE AGGREGATE PRODUCTION PLANNING PROBLEM

 What is this problem?
given a set of forecasts of demand for products produced by company, what are ideal levels of
capacity, production, subcontracting, inventory, stockouts, pricing over a specified time
horizon?
 Goal: satisfy demand while maximizing profit (or any other objective)
 often there are numerous constraints that are binding
General Considerations
 Manager must identify specific operational parameters for every period

e.g. production rate, workforce, overtime, machine capacity, subcontracting, backlog, inventory on
hand
 How long should the planning interval be?
the further into future, the less accurate forecast
 Changing the “givens” - use of promotions, expanding product portfolio, sharing equipment
 relevant cost elements must be made available
Complexity
 Complexity of aggregate production planning due to no constant demand

 How meet demand?
o Use inventory to absorb fluctuations
o Vary the size of workforce
o Keep the size of workforce constant but change production rate
o Changing overtime parameter
o Using shutdowns during low demand season
o Backlog and subcontract excess demand
o Reject some demand items during peak season
10.2 THE COST COMPONENTS

 Production costs
o Labour/ regular prime cost (typically linear; increase with production rate)
o Overtime cost (appears at production above certain production rate; U-shapedfirst only bottleneck
overtime, later more operation must run on premium ratesefficiency of workers decreases)
o Capacity change cost (hiring/ laying; usually cheaper to squeeze than to expand; U-shaped)
o Subcontracting cost
o Production rate change cost
 Inventory associated costs (holding cost)
 Cost of insufficient capacity (shortage cost)
 Control system cost
Along with costs we also face constraints
 Typical constraints that must be considered include

o Limits of overtime
o Number of layoffs
o Capital
47
o Limits of subcontracting
o Limited space of inventory
o Limits on changing production rate
10.3 LEVEL AND CHASE STRATEGIES

 Production manager can follow two pure strategies:
Level Strategy
 Maintain constant workforce size and produce similar quantities each period
 Use inventories and back-orders to absorb demand peaks and valleys
 inventory might become too big high Cc
Chase Strategy
 Minimize finished goods inventories by trying to keep pace with demand fluctuations
 Matches demand (seasonal fluctuation) by varying either work force level or output rate
 risk of missing demand and thus occurring
Mixed strategies often applied
 imply more flexibility and thus seen better from a cost minimization point of view
10.4 AN LP FORMULATION OF THE PROBLEM

 Provably optimal solution Methods, such as Linear Programming
 Linear objective function to be mini-/ maximized w.r.t. objective variables
Linear Programming Model
Advantages Disadvantages
48
 LP models provide much better solution  Requirement of linear cost functions

compared to “feasible solution methods  Reluctance to accept from management
 Results can be obtained fast even for the
big (100,000 variables and constraints)
problems
 Constraints can be easily introduced
 Sensitivity analysis available due to dual
solution
10.5 COMPARISON OF THE METHODS; ADVANTAGES AND DISADVANTAGES

10.5.1 Graphic–Tabular Method
10.5.1 Graphic–Tabular Method
 Easy to use and apply  …not mathematically optimal
 quite practical when developing them  small objective function (no guarantee
on a spreadsheet how far the solution is from optimality!!
 Always guarantees feasible solution
with given demand but…

10.5.2 Mixed Integer Linear Programming Model (MIP)
Mixed Integer Linear Programming Model
 Almost unlimited modelling flexibility  Even harder to understand by
 Not only linear functions can be used management
 Good results can still be obtained even for  Much more complicated theory behind
big problems (many variables/ constraints BUT with more advanced technology,
problems) implementation becomes more accept and
 Solutions are obtained with some sort of used
optimality guarantee, the lower bound for
the minimum cost is part of the algorithm
output
 constraints can be easily introduced
49
11 MATERIAL REQUIREMENTS PLANNING (MRP). LINKS BETWEEN THE AGGREGATE

PRODUCTION PLANNING PROBLEM, THE MASTER PRODUCTION SCHEDULE, AND MRP. BOM
AND LEVEL CODING. INFORMATION REQUIRED FOR MRP. THE GENERAL MRP APPROACH.
STRENGTHS AND WEAKNESSES OF MRP. (CHAPTER 15)
 Previously, optimal order strategies for single products and demands only
 Now, production comes into play with subcomponents and Materials!!
 e.g. Three assemblies with uniform demand all having one subassembly bull-whip affect
11.1 LINKS BETWEEN THE AGGREGATE PRODUCTION PLANNING PROBLEM, THE MASTER
PRODUCTION SCHEDULE, AND MRP
Sequence
1. medium-range aggregate production planning (LP)
 concerned with establishing production rates, work force sizes, and inventory levels for
something on the order of 6–24 months into the future
2. Master Production Scheduling (MPS)
 central component of the whole framework
primary interface between marketing and production
 takes aggregate production plan (with implied constraints) and disaggregates it into a production
schedule of specific products to be produced in each facility
3. Materials Requirements Planning (MRP)
 explodes master production schedule into implied, detailed production/procurement
schedules/runs
 heart of all this is the forecast of demand (sales/marketing input)
 Closed-Loop MRP as an extension that fills up gap of MRP which did not use capacity check
/input from other departments
o What are we going to make? (using the forecast)
o What does it take to make it? (using resource requirements and BoM)
o What do we have? (using inventory records)
o What do we need, and when? (using MPS)
Weaknesses of old inventory systems

 assumption of independent demand for each item and decision based on that, leading to
1. No need to forecast once production plan is set up (dependant demand; simple arithmetic)
2. SS was based on smooth demand
unrealistic for components due to batch production
3. Aimed to replenish stock following large demand
 components demand can be very erratic with periods of inactivity
4. Several components needed for one assembly
 inventories should not be treated individually
(e.g. final product needs 15 components with each demand satisfied demand with 96%  probability to assemble product is
0.9615 = 54%)
11.2 BOM AND LEVEL CODING

11.2.1 BoM
 Essential for MRP and ERP
 listing of all subassemblies, intermediates, parts, and raw materials that go into a parent
assembly, showing quantity of each required to make an assembly
 BoM is used to project needs w.r.t. components for a production lot of a particular assembly or
subassembly
50
 Modular form of BoM of a particular SKU shows all of its immediate components and their
number per unit of the parent
Figure 3: BoM red toy car
11.2.2 Level Coding

 Provide systematic framework for exploding back the implications on all components of a given MPS
Each item (or equivalently BoM) is assigned a level code
Level 0: finished product (or end-item) not used as a component of any other product
Level 1: direct component of level 0 item; level 1 item could also be finished product in itself
Level 2: direct component of level 1 item; level 2 item could also be finished product in itself or
finished product
Level n: direct component of a level (n − 1) item, i.e. a component on a bill of materials with level
code (n − 1)
 MRP avoids double usage by always selecting higher level number to code item
Figure 4: Level-by-level coding for the red toy car
Indented form of BoM: there is a one-to-one correspondence between the indentation and the level code
 car is sold in three forms, a car and truck gift pack, a multicar pack, and as an individual item (regular
51
external demand).
11.2.3 Lead Time (Offsetting) and Routing

Lead Time
 time that elapses from when we decide to send PO/ producing until moment when material is
physically present for the first processing operation
Routing
 routing shows sequence of production operations and standard hours for each operation
11.3 INFORMATION REQUIRED FOR MRP

 Information required to carry out MRP:
1. Master Production Schedule (MPS) projected to planning horizon
2. Forecasts of demand for components, subject to direct customer demand, up to planning
horizon
3. Inventory status of each item (including backorders)
4. Timing and quantities involved in any outstanding order
5. All relevant BoMs and associated level codes
6. Sequences of production operations and associated work centres along with standard hours
for each operation (for capacity planning)
7. Production or procurement lead times
11.4 THE GENERAL MRP APPROACH. STRENGTHS AND WEAKNESSES OF MRP
52
12 SHORT-RANGE PRODUCTION SCHEDULING. PERFORMANCE MEASURES. GANTT-CHARTS.

HORIZONTAL VS. VERTICAL LOADING. SEQUENCING RULES. JOHNSON’S RULE FOR THE 2-
MACHINE CASE. (CHAPTER 16)
53

Supply Chain Inventory Production Planning Exam Prep Oral

Uploaded by

Copyright:

Available Formats

Supply Chain Inventory Production Planning Exam Prep Oral

Uploaded by

Document Information

Original Description:

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

Supply Chain Inventory Production Planning Exam Prep Oral

Uploaded by

Copyright:

Available Formats

What are the main topics covered in the document?

What are the main topics covered in the document?

What are some forecasting techniques discussed?

What are some forecasting techniques discussed?

SEMESTER 1- SUPPLY CHAIN INVENTORY & PRODUCTION PLANNING

 mixed (additive and multiplicative)

1.2 MOVING AVERAGE AND FORECASTING (MEDIUM RANGE)

1.3.1 Exponential for Level Model (smoothed average)

 intuition is to put less weight on older data

1.3.2 Exponential for Trend Model

 these are then updated based on new data

1.3.3 Exponential for Seasonal Model

 Alpha, beta and gamma..

1.4 MEASURES OF FORECAST ERRORS.

Mean Square Error (MSE) (Variance)

 When updating the MSE, one should use exponential smoothing

 MSEt requires less storage of data (only most recent used)

Mean Absolute Deviation (MAD)

Mean Absolute Percentage Error (MAPE)

 good if one doesn’t know an item’s typical demand volume

2 ORDER QUANTITIES WHEN DEMAND IS APPROXIMATELY LEVEL. GENERAL ASSUMPTIONS,

8. Time variable of notations should be the same (convert accordingly)

2.2 EOQ MODEL (ECONOMIC ORDER QUANTITY)

2.3 EPQ MODEL

 Assumption Replenishment quantity (Q) arrives at a rate of m units per time

2.4 EOQ WITH QUANTITY DISCOUNT

finding the best Q to reduce the costs accordingly

2. Compare EOQ(d) with Qb:

b. EOQ(d) < Qb, then do step 3

3.1 GENERAL ASSUMPTIONS

10. Time variable of notations should be the same (convert accordingly)

3.2 THE WAGNER-WHITIN-METHOD

3.3 THE SILVER-MEAL HEURISTIC

 order costs + carrying 2nd periods demand one period

3.4 OTHER PROCEDURES

3.4.2 EOQ Model

3.4.3 Least Unit Cost (LUC)

3.4.4 Part-Period Balancing (PPB)

 rather order 60 units than170

result rounded to nearest higher integer

3.5 WHEN TO USE HEURISTICS

Key issues to solve:

 Manager must consider following questions to answer the previous three:

3. Inventory Position/ Available Stock

on-order: already ordered but not received yet

What, if demand can not be satisfied? Two extreme cases:

4.1 THE (S,Q)-SYSTEM

 Fixed Q is ordered when inventory drops to reorder point s or lower

4.2 THE (S, S)-SYSTEM

4.3 THE (R, S)-SYSTEM

4.4 THE (R, S, S)-SYSTEM

4.5 COST AND SERVICE OBJECTIVES

3. SSs based on Customer Service

4.5.1 Simple-Minded Approach

4.5.2 Minimizing Costs

4.5.3 Customer Service

4.5.4 Aggregate Considerations

4.6 HOW TO CHOOSE S IN (S,Q)-SYSTEM WITH A GIVEN Q

6. (Optimal) order size Q is predetermined (we focus on B-class items)

If nis exactly an integer, go to Step 4 with n = n