Push Vs.

Pull
Part Two – Focus on MRP

IE 3265 POM
R. R. Lindeke, Ph. D.
Components of a Pull System

 Master Production Schedule – a breakout of

the Aggregate Plan
 Uses forecast demands/Firm Customer Orders/Safety
Stock Levels/Internal Orders
 Actual MRP System
 Structured Bill of Materials
 Production Times
 Detailed Job Shop Schedules
 Detailed Inventory Records
Defining Explosion Calculus

 This is the crux of MRP

 MRP converts “Dependent Demands” derived from
the Zero Level of an ordered product into a series of
Job Orders for:
 Customer Orders
 Internal Orders
 Safety Stocks to meet unexpected demands
 The MRP explodes each order of “0’s” into the
required series of Children orders needed internally
to build the required products
Some Other Important Terms

 Gross Requirements:
 Total needs of a part for ALL parents in a MPS & safety stock
along with internal needs – like for maintenance parts,
replacement parts, etc
 Scheduled Receipts:
 These are open orders placed but not yet received
 Planned Receipts:
 Planning done over time horizon to keep on hand inventories
from dropping below zero (safety stock level) which are
backed up into orders to arrive when the shortage would
have occurred
Some Other Important Terms, cont.

 Planned Order Releases:

 A release when an order for a specified quantity (EOQ,
L4L, etc) is done
 This should happen in the correct “Time Bucket” – a
time bucket is the time stage used during projection
 Order release considers: Setup; Processing time;
Handling time; Delays, curing, waiting times
 These apply to each item in the product string
 We will consider a Seat Assembly for a Ladder Back
Chair
Looking at a Ladder Back Chair:
Structured Bill of Materials for Ladder
Back Chair
Focusing on the Seat Assembly

 We can use several Order Schemes:

 EOQ or other fixed order quantity (FOQ) schemes
 POQ periodic order quantities which will order a
computed amount needed to cover orders over an
extended (specified) period of time. The orders are
released at fixed time intervals corresponding to the
time buckets
 Lot-for-Lot (L4L) – a simple plan that puts out a “time
bucket” order to match lot demand – simple but not
necessarily the most economical
Planning for the Seat Ass’bly
 Used in two parent products: Ladder back and Kitchen Chairs
 Requirements:
– Wk 1: 150 LBC
– Wk 4: 120 KC
– Wk 6: 150 LBC
– Wk 7: 120 KC
 Lets try it as a L4L plan – zero on hand inventory, no safety
stocks
 Lead time for Chair assembly is 2 weeks
 Assume we have 37 in inventory from previous plans and
expect a delivery of 230 as we enter this period
 L4L Plan (manual MRP!)

1 2 3 4 5 6 7 8 9
Gross
150 120 150 120
Re’ment:

Sch.
230
Receipts
37 +
Proj.
80 = 117 117 0 0 0 0 0 0
Inventory
117
Planned
3 150 120
Receipts
Planned
O. 3 150 120
Release
Using a (3 period) Periodic O.Q.

1 2 3 4 5 6 7 8 9
Gross
Re’ment: 150 120 150 120
Sch.
Receipts 230
Proj. 37+80
Inventory = 117 117 117 150 150 0 0 0 0
Planned
Receipts 153 120
Planned
O. 153 120
Release
Focusing on the “Periods”

 Period 1 (1-2-3):
 needs are 150
 Receipts/onhands are 230 + 37
 Xs is 117 which enters inventory
 Period 2 (4-5-6):
 needs are (120 + 150) = 270 units
 Less onhand (117) means we need 153 units for period
to arrive at start of period
 Release order in period two to arrive by period 4 – start
of second period
Final Period:

 Period 3 (7-8-9):
 Needs are: 120 units
 No on-hand inventory is carried forward so must release
an order in period 5 (for 120) to arrive in period 7 when
needed
 Next, lets look at a deeper need:
– The frame parts of the seat Ass’bly which has a 1
week lead time (each seat ass’bly requires 4
frame pieces)
Try it with L4L
1 2 3 4 5 6 7 8 9
Gross
150 120 150 120
Re’ment:
Sch.
230
Receipts
37 +
Proj.
80 = 117 117 0 0 0 0 0 0
Inventory
117
Planned
3 150 120
Receipts
Planned
O. Rel 3 150 120
(SA)
POR
12 600 480
Frames
Effect on POQ method

1 2 3 4 5 6 7 8 9
Gross
Req’men: 150 120 150 120
Sch.
Receipts 230
Proj. 37+80
Inventory = 117 117 117 150 150 0 0 0 0
Planned
Receipts 153 120
Planned
O. Rel 153 120
(SA)
P. O. R.
frames 612 480
Costing the plans -- Basis is Setup vs.
Holding (as expected!)

 Given: $0.40/ass’bly holding; $120 Setup

– L4L on SA:
 .4*(117*3) + 3*120 = 140.4 + 360 = $500.40
– POQ:
 .4*(3*117 + 2*150) + 2*120 = 260.4 + 120 = $500.4
 Given: $0.10/ass’bly holding; same setup
– L4L on SA:
 35.10 + 360 = $395.10
– POQ:
 65.10 + 240 = $305.10
Continuing but adding in Frame
Ass’blies

 For Frames: K = $100; h = $0.02/piece

 L4L:
 SA Costs + (1092*.02) + 3*100 = 395.10 + 21.84
+ 300 = $716.94
 POQ:
 SA Costs + .02(612 + 480) + 2*100 = 305.10 +
21.84 + 200 = $526.94
Using EOQ (previously found to be
230) Recognizing
shortfalls in P4 and
P7 we release EOQ
orders to fill needs

1 2 3 4 5 6 7 8 9
Gross
Req’men: 150 120 150 120
Sch.
Receipts 230
Proj. 37+80
Inventory = 117 117 117 227 227 77 187 187 187
Planned
Receipts 230 230
Planned
O. 230 230
Release
Comparing Methods

 Compute Average Inventory (period)

– Here:
 L4L is (3*117)/9 = 29
 POQ is (3*117 + 2*150)/9 = 72.3
 EOQ is (3*117 + 2*227 + 77 +3*187)/9 = 160.3

 The choice is based on comparing setup

costs vs. the average inventory costs
Statement of the Lot Sizing Problem –
looking for an optimal

 Assume there is a known set of requirements (r1,

r2, . . . rn) over an n period planning horizon. Both
the set up cost, K, and the holding cost, h, are
given.
 The objective is to determine production quantities
(y1, y2, . . ., yn) to meet the requirements at
minimum cost. The feasibility condition to assure
there are no stockouts in any period is:
j j

 y  r
i 1
i
i 1
i for 1  j  n
Methods

 Property of the optimal solution: every optimal solution orders

exact requirements: that is,

y1  r1 or y1  r1  r2 ,. . ., or y1  r1  r2  ...  rn

 One method that utilizes this property is the Silver Meal

Heuristic. The method requires computing the average cost for
an order horizon of j periods for j = 1, 2, 3, etc. and stopping at
the first instance when the average cost function increases. The
average cost for a production quantity spanning j periods, C(j),
is given by:

C ( j )  ( K  hr2  2hr3  ...  ( j  1)hrj ) / j

Methods (continued)

 Another method that is popular in practice is part

period balancing. Here one chooses the order
horizon to most closely balance the total holding cost
with the set-up cost.
 Finally, a third heuristic is known as the least unit
cost heuristic. Here one minimizes the average cost
per unit of demand (as opposed to the average cost
per period as is done in the Silver Meal heuristic.)
The average cost per unit of demand over j periods
is given by:
C ( j )  ( K  hr2  2hr3  ...  ( j  1)hrj ) /(r1  r2  ...  rj )
Methods (concluded)

 Experimental evidence seems to favor the

Silver Meal Heuristic as the most cost
efficient among the four discussed in the text.
 Optimal lot sizes can be found by using
backwards dynamic programming.
 The (best?) heuristic method for lot sizing
subject to capacity constraints is shown next
Lets Try Silver-Meal (assumes K is
$140; holding/unit/period is $0.10)

 Over our 9 period time horizon (looking at Seat Ass’bly

Demands)
 R: (120, 0, 0, 120, 0, 150, 120, 0, 0)
 C(1) = K = 140
 C(2) = (K + .1*0)/2 = 70
 C(3) = (K + .1*0 + 2*.1*0)/3 = 46.67
 C(4) = (K + .1*0 + 2*.1*0 + 3*.1*120)/4 = 44
 C(5) = (K + .1*0 + 2*.1*0 + 3*.1*120 + 4*.1*0)/5 = 35.20
 C(6) = (K + .1*0 + 2*.1*0 + 3*.1*120 + 4*.1*0 + 5*.1*150) =
41.83
 Since C(6) > C(5) Stop. 1st order release is (R1 .. R5) =
240 units but due to lead time we must release it in period
-2 (before the current time horizon)!!!
Continuing
 Reset clock to Period 6 as new Start Point:
 C(1)6 = K = 140
 C(2)7 = (K + .1*120)/2 = 76
 C(3)8 = (K + .1*120 + 2*.1*0)/3 = 50.67
 C(4)9 = (K + .1*120 + 2*.1*0 + 3*.1*0)/4 = 38
 Done with plan – so make a lot (in period 4 due to lead
time) for arrive in period 6 for 270 units.

 Notice: each time the cost inverts:

 we reset the clock and trigger an order release of the
amount required by the sum of requirements identified
before the cost inverted
One Last Topic of Interest

 Capacity Constraints
– Built from a comparison of:
 R the requirements vector
 C the Capacity vector
 Model is feasible only if: (Ci) >= (Ri)
 Lets examine a simple implication:
 R: (17, 37, 92, 55, 80): (Ri) = 281
 C: (60, 60, 60, 60, 60): (Ci) = 300
 This is a feasible plan – BUT –
 It is obvious that we could be short in period 3 and 5 without
some smart planning!
 Lets Try to Be Smart! (Initial Thought –
lets try L4L)

1 2 3 4 5

Gross
17 37 92 55 80
Req’r

Planned
17 37 60  55 60 
Rec’pt

Xess
43 23 -32  5 -20 
Cap
Fixing the Shortfalls – push back
needs and build inventory!

 Move P3’s shortfall back in time

– 23 to P2 (which fills capacity of P2)
– 9 additional ones back to P1
 Move P5’s shortfall back too
– 5 to P4 (filling capacity of P4)
– the other 15 all the way back to P1
– After redistribution: P1 Production requirements are
9 + 15 + 17 = 41
Revised ‘Schedule’

1 2 3 4 5

Gross
17 37 92 55 80
Req’r
Planned
41 60 60 60 60
Rec’pt

Inventory 24 47 15  20 0
Shortcomings of MRP

 Uncertainty. MRP ignores demand uncertainty, supply

uncertainty, and internal uncertainties that arise in the
manufacturing process.
 Capacity Planning. Basic MRP does not take capacity
constraints into account.
 Rolling Horizons. MRP is treated as a static system with
a fixed horizon of n periods. The choice of n is arbitrary and
can affect the results.
 Lead Times Dependent on Lot Sizes. In MRP lead
times are assumed fixed, but they clearly depend on the
size of the lot required.
Shortcomings of MRP, cont.

 Quality Problems. Defective items can destroy the

linking of the levels in an MRP system.
 Data Integrity. Real MRP systems are big
(perhaps more than 20 levels deep) and the integrity
of the data can be a serious problem.
 Order Pegging. A single component may be used
in multiple end items, and each lot must then be
pegged to the appropriate item.