HD28
.M414
Dewey
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
^A
BEHAVIORAL SIMULATION MODEL OF
FORECASTING AND PRODUCTION SCHEDULING
IN A DATACOMMUNICATIONS COMPANY,^
by
John D.W. Morecroft
WP-1766-86
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
D-3807
^A
BEHAVIORAL SIMULATION MDDEL OF
FDRECASTING AND PRDDOCTION SCHEDULING
IN A DMACOMMUNICATIONS COMPANY
^
by
Jolin D.W.
Morecroft
WP-1766-86
p|L((l(
KID nfoCD^SCi'
D-3807
A BEHAVIORAL SIMULATION MODEL OF
FORECASTING AND PRODUCTION SCHEDUUNG
IN A DATACOMMUNICATIONS COMPANY
by
John D.W. Morecroft
System Dynamics Group
Sloan School of Management
Massachusetts Institute of Technology
Cambridge MA
March 1986
D-3807
A BEHAVIORAL SIMULATION MODEL OF
FORECASTING AND PRODUCTION SCHEDUUNG
IN A DATACOMMUNICATIONS COMPANY
Abstract
The paper describes a behavioral
simulation model of forecasting and
a datacommunications company. The model allows
production scheduling in
you to think about the different 'players'
whose choices and
actions
regulate orders and production. There are business planners (who provide
forecasts), factory schedulers, expediters, customers and account
executives. You step into their shoes. You examine their responsibilities,
their goals and incentives and the sources of information that attract
their attention
--
all
with the intention of understanding the logic behind
choices and actions. Then you stand back from the detail and, with
the help of diagrams and simulations, you explore how the players interact
and cooperate, and how the factory balances supply and demand.
their
D-3807
INSIDE THE
FACTORY
Imagine youself as the manager
of
a factory that produces workstations,
PBXs, key systems and other datacommunications equipment. Your factory
has been
criticized recently for
slow deliveries. The delivery interval
much
workstations ranges from six months to one year,
The problem
competitors.
is
not improving.
of
longer than major
You begin searching
for
an
explanation.
Many
ideas
come
due
intervals are
to mind.
particularly
But you
in
could argue that the factory's long delivery
to errors in the sales forecast.
the factory better and
forecasts.
One
make
deliveries on time
realize
that
forecasts
if
You could obviously
run
you were given accurate
never be accurate,
will
the datacommunications market where technology
is
causing products and prices to change rapidly. Alternatively, one could
argue that the factory's long delivery intervals are due
suppliers
who
needs more.
convincing.
can't expedite parts
on
But,
Suppose
reflection,
for
example
materials arrives unexpectedly,
What would you do
and materials
even
this
with workstations,
their
On average
machines are
truck.
of parts
months
and
extra workstations.
Let's
a workstation
this
for delivery, but
now
to receive
they've waited say eight
in
not entirely
where would you send them?
say there are 100 customers scheduled
month.
is
shipment
make 50
to
inflexible
the factory suddenly
argument
that a large
enough
if
to
the shipping dock, ready to be loaded onto a delivery
Suddenly, 50 more workstations become available. Curiously,
D-3807
despite the eight month delivery interval, the factory cannot ship the 50
show another 50
extra workstations, because the order book doesn't
customers who are expecting
it
to receive
a workstation
shows 100 customers who are expecting
several hundred customers
twelve months.
early
and
It's
who
field
month. Instead
delivery this month,
and
are expecting delivery during the next
not a simple matter to ship the extra workstations
customers must be informed
--
this
of early deliveries,
so must the sales
support organization. Moreover, the factory has to adjust the
production schedule to account for early deliveries.
As you
(since
think about
it,
the factory's problem
most customers receive
but, instead, with
slow deliveries
but the agreed delivery interval
How does
their
is
--
is
not with late deliveries
workstations on the promised date)
workstations are shipped on time,
too long!!
You seem
to
be gridlocked.
the factory get out of this delivery situation, and, perhaps more
important,
how does
it
avoid extended delivery intervals
in
the
first
place?
To
help explore these questions a simulation model has been developed of a
factory,
customers and account executives. The model allows you
about the different
'players'
whose
to think
actions and choices regulate orders and
production. There are business planners (who provide forecasts), factory
schedulers, expediters, customers and account executives.
their
shoes.
You examine
their responsibilities, their
You
step into
goals and incentives
D-3807
and the sources
of information that attract their attention
intention of understanding the logic behind their choices
you stand back from the
detail
how
simulations, you explore
the factory balances supply
and
--
all
with the
Then
actions.
and, with the help of diagrams and
the players interact and cooperate, and
how
and demand.
Production Planning and Factory Scheduling
Consider
production planning, which
first
is
a multilayer process that
consolidates information from forecasters, from product managers and
from the factory. Figure
shows
1
the information entering the production
planning policy.
An important
input
is
the market forecast
based on the marketing
business plan. The business plan starts from an estimate
volume
for
all
staff's
of total industry
classes of datacommunications equipment. From historical
data and various business assumptions (new product introductions, price
changes, competitor actions, expected delivery
compute the company's expected share
multiply industry
of industry sales.
volume and expected share
sales forecast by product
forecasting process
is
line
intervals), forecasters
to
generate the company's
over a two-year planning horizon. The
a logical exercise
in
gathering and processing
market information. But as the forecast enters the factory,
production scheduling,
for
accuracy
in
it
is
Then, they
modified by
its
own
to
be used
in
credibility (its reputation
the factory), by pressures from inventory, delivery interval
D-3807
Perceived
Marketing
Forecast
Delivery Interval
Credibility of
Forecast
Scheduled
Production
Measured
Backlog
Adequacy
of
Inventory
Backlog
Inventory
Tightness of
Inventory Policy
^-3Z3'=\B
Figure
1:
Production Planning
D-3807
and backlog.
identify the
Let's consider
each
of
people most responsible
these factors
for
in
turn
in
order to
modifying the forecast and their
rationale for doing so.
Let's
begin
with
the
forecast
datacommunications companies
the factories.
market
frequently,
for
The reasons are easy
is in flux.
New
credibility.
to
market forecasts
understand.
forecasters simply don't
know
in
to lack credibility at
The datacommunications
daily,
prices
change
Factory managers feel that
spring up.
the sales potential of a given product line
(despite their elaborate forecasting methods)
be
common
quite
is
products are introduced almost
new competitors
often pared-down, to
It
'on the safe side',
and therefore, forecasts are
and
particularly to avoid the
very visible inventory costs resulting from excess production.
What
if
error,
what source
the forecast
is
inaccurate?
of information
How does
the factory
would reveal the
error,
know
of the
who would be
responsible for adjusting the production schedule and would they have the
incentives
Now we
shoes
and appropriate information
to
make
the correct adjustment?
are getting into the heart of factory scheduling, stepping into the
of the schedulers themselves.
The production schedule
can envisage a
is
defined over a rolling six-month horizon.
One
planning chart, with boxes marked across a page. The
extreme left-hand box shows the number
of units of
a given product-line
D-3807
8
(say workstations) scheduled for production
the schedule
may be expressed
difference for this discussion).
number
of units
scheduled
for
in
weeks
in
or
the current month
even days, but
it
(in
fact
makes no
The extreme right-hand box shows
the
The boxes
production six months from now.
in-between show monthly scheduled production for months 2 through
Each month the planning chart
is
updated, the current month's box "drops
and a new box
off the left-hand side of the chart,
right-hand side, rather
in
the
new box
is
like
The scheduler assigns incoming customer orders
time.
to
A new customer
is
due
The scheduler looks
month
3.
He
finds
order
for installation
in
is
on the
in'
The number
months hence.
to the appropriate box.
understand the process, imagine that the factory
workstation) which
'rolls
the steps on a moving escalator.
the forecast of customer orders six
three month interval.
5.
is
trying to deliver
To
on a
received (say an order for a
and shipment
in
three months
the planning chart for the box corresponding
a workstation that
assigns the customer order to
it.
is
scheduled
for production
That particular workstation
is
and
now
'earmarked' for a customer and cannot be assigned again.
The process works smoothly as
But what happens
deliver on
when
a three month
have been scheduled
150 workstations
in
long as the original forecast
the forecast
interval.
is
was
low? Again the factory
Based on the
forecast,
1
accurate.
is
trying to
00 workstations
the time slot three months hence. But orders for
arrive.
The scheduler assigns the
first
100 orders
D-3807
exactly as before,
until
the entire box
is
assigned. Then he assigns the
remaining 'unexpected' orders to the next available box
to
a
later
time
-- in
other words
Having done so, he then informs the salesforce that
slot.
workstations are available only on an extended delivery interval. Because
of the long production planning horizon,
an open production
slot for
a new customer order.
has been assigned a production
who
schedulers,
the
first
people
no incentive
slot,
When a customer
done
the scheduler has
order
his job.
So
are 'close' to the customer order backlog, and therefore
in
company
the
argue
to
schedulers can almost always find
to
see when the forecast
for higher production,
too low, have
is
as long as production
slots
remain open.
The same
production
works
in
scheduled
in
logic
is
find early time slots for
reverse.
excess
of
When
it
to
new customer
has done
an open production
his job,
wrong amount
(too
to
will
eventually
production.
in
in
come from
,
to
two months hence. The scheduler
to
is
producing the
argue
for
a lower
case, pressure to reduce the production
finished inventory.
If
customers are unable
then products awaiting shipment
finished inventory
and
three months, but finds he can
much) he has no incentive
take an early delivery
accumulate
slot only
in this
optimistic
So a scheduler might
orders.
so again, although he knows the factory
production plan. However,
plan
is
customer orders, schedulers begin
receive an order for a workstation due
assign
the forecast
and signal a factory manager
will
to curtail
D-3807
10
Customer Ordering
Figure 2
shows the influences on customer
products before they
aware
Let's
suppose
become
for the
given product line
of
will
the
even consider purchase. Then, once the customer
of the product, traditional factors
delivery interval
is
because customers must learn about the company's
principal influence,
is
ordering. Sales effort
is
such as price/performance and
important.
sake
of simplicity that the
price/performance of a
constant, and that sales effort (which you can think
as the number of hours per month the salesforce spends with
customers)
is
constant.
How do customer
orders vary with changes
in
delivery interval? Think about this question from the perspective of
account executives as they contact potential customers. Suppose that
account executives have been
delivery interval.
be extended
on the
who
is
five
Then
to five
month
willing
selling workstations
factory schedulers
announce
that the interval
months. Account executives can
interval, but
to wait the
they
will
customer orders increase as the
still
sell
must
workstations
take longer to find a customer
extra two months.
customer orders decline as the
on a three month
interval
rises,
interval falls.
The
net effect
is
that
and conversely,
that
11
D-3807
Sales Effort
Orders Booked
Perceived
Reliability of
Delivery
Perceived
Delivery Interval
Perceived
Price / Performance
Ratio
^-^Z3tB
Figure 2: Customer Ordering
D-3807
12
DYNAMICS OF PRODUCTION SCHEDUUNG AND CUSTOMER ORDERING
If
one accepts
described above, then
first
planners and schedulers behave as
that customers,
how do they
interact?
To answer
this
visualize the feedback loops connecting the 'players'.
question
let's
There are two
important loops.
shows a negative feedback loop connecting customers' ordering
Figure 3
the company's deliveries. Consider the loop's operation
becomes
available
less
(the
shuts
factory
if
to
product suddenly
down,
or
there
a
is
transportation strike). Shipments decline, causing delivery interval to
rise.
to
When
make a
and the
the delivery interval rises
sale,
total
so
that, for
number
demand mechanism
it
takes more time for salespeople
any given sales
of orders
that brings
booked
effort,
stabilizes.
customer orders
customer orders decline
Here
into
is
a self-regulating
an exact balance with
shipments.
shows a
Figure 4
positive
feedback loop connecting cutomers' ordering
to
forecasting, production scheduling
and shipping. This loop can
generate
as the following argument shows.
Suppose
A
self-fulfilling
that production declines
decline
factory
forecasts,
in
due
to
a previously inaccurate forecast.
production quickly curtails shipments (assuming that the
keeps very
little
finished inventory).
delivery interval rises, time per sale increases
When
orders
readily
fall,
When shipments
fall,
and customer orders
the
fall.
market share declines, and so eventually does expected
D-3807
13
Sales Effort
Customer Orders
Orders Booked
Time Per Sale
Shipments
Delivery
Interval
Product
Availability
Figure 3: Negative Feedback Loop Connecting Customers'
Ordering
to
Shipments and Delivery
Interval
D-3807
14
Industry
Sales
Current
Market
Customer
Orders
Shore
Expected
Estimoted
Market
Industry
Stiare
Time Per Sale
Delivery Interval
Volume
Marketing
Forecast
Discount for
Forecast
Credibility
Shipments
Manufacturing
Schedule
/
/
/
/
/
Product
Availability
/
Production
y'
Tightness of
Inventory Policy
Accuracy of
Forecast
-*-
k- 51+lB
Figure 4: Positive Feedback Loop Connecting Customers'
Ordering to Forecasting, Scheduling and Shipping
D-3807
15
market share, as
share
it
is
based
largely
on
the marketing forecast
falls,
is
historical data.
therefore feeds back to reinforce
In
reduced and so too
further
The
production schedule and the production rate.
Because expected
initial
product shortage
itself.
We
the real system the two loops are combined.
can use simulation
understand how they operate. Figure 5 shows the system's response
one time twenty percent increase
that, in the factory
model, sales
free to set sales effort at
in
effort is
'exogenous'
--
any value he thinks appropriate.
in
D-3806, that
a datacommunications company can vary widely depending
compensation scheme. So a simulation experiment
assumes a
large,
twenty percent, increase
The model
and production
(-1-)
(-2-) of
at
100
units per
month.
of
110
units per
a month, but only
(-2-)
increases slowly and steadily
to
(-1-)
sales effort increases,
But
month
half the possible increase.
until
it
--
eroded
until
an increase
is
they
of
10
Meanwhile production
exactly equals customer
why are orders permanently depressed? As
shows, the immediate cause
months
When
certainly
is
customer orders
by, orders (-1-) are gradually
a new equilibrium
(-1-).
sales effort
that
increase correspondingly to a peak of 120 units per month.
units
orders
in
starts in equilibrium, with
However, as time goes
settle
is
we
case,
In this
of the
orders
a
to
the modeler
on the terms
plausible).
to
sales effort (the reader should note
in
know, from the sales planning and control model described
sales effort
the
is
figure 6
delivery interval (-1-) which rises from 3
a permanently higher value
of
3.75 months. Fewer customers are
D-3807
16
2p
CO
1
3 rco
130.000
120.000
no
000
100.000
90 000
40.000
Time
Figure 5: Customer Orders and Production
in
1
2}
2
pdi
Sales
-
20% Step
Effort
idi
4.500
4000
3,500
3.000
^}
2 500
40 000
Figure 6: Delivery Interval
--
20%
Step
in
Sales
Effort
D-3807
17
willing to
But
why
place an order
when
doesn't the interval
orders has passed?
the interval
back
fall
To answer
this
is
extended.
to three
question
months once the surge
we need
of
to look closely into
the 'mechanics' of forecasting and scheduling. Figure 7 provides part of the
story.
When
orders increase unexpectedly, the forecast increases too, but
only gradually.
interval),
As customer orders
the forecast
(-1-)
if
decline (due to increased delivery
changes course. Instead
toward reference customer orders
would achieve
(-2-)
delivery interval
(-3-),
were
of continuing
(the value that
fixed at three
slowly on the final and depressed value of orders.
forecast
is
production
self-fulfilling.
in
excess
factory schedulers to
The only way out
of the forecast.
do
so. Figure 8
customer orders
months)
In
it
is
homes
schedule
no pressure on the
shows why. When customer orders are
unexpectedly high, the schedulers simply assign the excess orders
earliest
surge
convenient
of orders
is
slot in the
(-1-)
--
yet,
-
higher than they should
at the
customers are
same
As a
if
result,
orders booked
(-2-)
rise
deliveries are to remain competitive
time, the illusion
satisfied,
of the
instead of expanding the schedule by adding
the excess orders to the forecast.
too high
to the
pre-planned manufacturing schedule. So the
absorbed by consuming a larger proportion
pre-planned schedule
in
other words, the
of this trap is to
But there
upward
exists
in
the factory that the
since each customer order
is
assigned a
production slot and scheduled for delivery on a date that the customer has
D-3807
18
3
2 CO
slbf
1
rco
130.000
1
2
2J
120,000
3J
N
4^
10.000
^
10
ID
100.000
90.000
-p-i
r—
20.000
10.000
0.0
1
1
1
1
p
40.000
30.000
Time
Figure 7: Self-Fulfilling Forecast
1
2
2
Tpoc
800
450 000
--
20%
Step
in
Sales
Effort
Ob
^
1
0,700
400.000
t
2
1
2
600
350 000
0.500
300.000
400
250.000
40 000
30 000
20.000
10.000
Time
Figure 8: Fraction of Planned Orders Committed and
Orders Booked
--
20%
Step
in
Sales
Effort
D-3807
agreed
1
The
to.
salesforce
illusion is easily
sustained, because the factory keeps the
on the earliest time slots available
up-to-date
the
in
who
production schedule, so salespeople tend to find customers
are
satisfied with the factory's schedule.
SELF-SUPPRESSING DEMAND
The
simulations above
policies
fail
show
that the factory's forecasting
-
regulate delivery interval
to
upward whenever the forecast
interval to drift
factory can inadvertently suppress orders for
understand how, consider the case
attracting
a growing proportion
of the sales planning
grow very
and
is
control
in six
As a
a new product
of
model
too low.
they allow the
in
Think back
line
which
To
is
D-3806. Customer orders can
months, not because customers
With such rapid salesforce-induced growth
forecast to be too low, not just for a
lines.
to simulations
are stampeding to buy, but instead because the salesforce
sell!
result, the
growing product
its
of sales effort.
They can double
quickly.
instead
and scheduling
week
it's
easy
is
anxious
to
for the sales
or a month, but for six
months
or a whole year. With a low forecast, the factory's scheduling policies
allow delivery interval to
upward
until
it
is
high
drift
enough
salespeople from allocating
Customer orders
will
upward. The interval
to deter
still
new product
line.
continue to
drift
customers from buying and
more time
therefore stop growing.
scheduling and salesforce time allocation
will
to
So
will
selling
the product.
the interaction of factory
suppress demand
for the
D-3807
20
POLICIES TO CONTROL DELIVERY INTERVAL
To prevent
manage
the problem of self-suppressing
the delivery intervals of
its
demand
the factory needs to
different product lines, to
ensure the
intervals remain competitive.
The idea
more than implementing a
special interval reduction program.
programs
senior
typically
of
managing
means
intervals here
Such
occur only when intervals are so high that they attract
management
By then the image
attention.
of
extended
intervals
is
already well-established with customers and with the salesforce, and
many
potential
orders have been
lost.
Managing
establishing a routine policy, at the level of schedulers
rapidly detects systematic bias
scheduling production
in
in
excess
control policies are described
intervals
means
and planners,
the forecast and compensates for
of the forecast.
Two
that
it
by
possible interval
and simulated below.
Monitoring and Control of Excess Orders
The
them
first
policy requires schedulers to monitor
to adjust the forecast.
is
helpful to explain.
is
and
now
delivering workstations with an interval of three
(competitive for the industry).
120 orders
for workstations,
In
due
the current
month the
for delivery three
scheduler looks at planned orders three months
He
Suppose
the
planning production of workstations over a six-month horizon
factory
is
An example
excess orders and use
finds only
100 workstations scheduled
factory receives
months hence. The
into the
for that
months
planning horizon.
month - 100 being the
forecast of orders generated three months ago, at the start of the planning
D-3807
21
process. With the present procedure the scheduler puts the 20 excess
orders into next month's time slot and then informs the salesforce of the
schedule change. With the proposed system the scheduler would,
addition, count-up
date
is
all
excess orders
greater than three months
number
to production planners.
excess orders
to result in
to the current
--
orders
(all
20
in
this
whose scheduled
Each month the planners would add the
month's forecast.
(
This procedure
excess customer orders and then the planners add 20
in
fact
delivery
case) and communicate the
double scheduling because the schedulers
the forecast. But
a customer order
is
first
may seem
assign the 20
units of production to
scheduled once, and once only,
by the scheduler. The planners use the 20 excess orders as a guide
increasing the forecast. Their concern
is
with planning the
future production, not with scheduling individual
The same procedure could
compensate
all
for
in
In this
of
customer orders).
case the scheduler would count-up
the time slots before
month three
that
assigned a customer order. He would then communicate
excess planned orders
volume
for
also be used to reduce the production plan to
a high forecast.
planned orders
in
to production planners.
have not been
this
number
The planners would
the excess orders from the current month's forecast.
of
subtract
22
D-3807
shows a
Figure 9
with
simulation of the order monitoring
a 20 percent increase
figure 9 with figure 5.
orders
sales
before,
peak however, orders
an equilibrium
increases slowly
of
until
(-1-)
sales effort increases, customer
decline only slightly, and then settle
120 units per month. Meanwhile, production
it
new
policy in operation, delivery interval
stays close to three months, despite the surge of orders.
no longer suppresses demand due
factory
improvement
in
(-2-)
exactly equals customer orders.
Figure 10 shows, that with the
(-1-)
The reader should compare
effort.
when
control policy,
increase correspondingly to a peak of 120 units per month.
(-1-)
After the
at
As
in
and
delivery performance occurs
to
extended
So
intervals.
the
The
because planners expand the
production schedule more quickly than the forecast alone would suggest.
Figure
(-1-)
1 1
is
shows what
is
happening. From month
steady at 100 units per month.
customer orders increase) the forecast
In
to
month
month 5
(-1-)
5,
the forecast
(the time
starts to
when
but only
rise,
gradually because forecasters do not anticipate (by assumption) the step
increase
in
forecast
(-1-)
sales effort that
has risen
to
is
causing the extra demand. By month 10, the
110
units per
10, forecasters are saying that, six
will
month
(in
months hence
other words,
-- in
month
in
month
16, orders
be 110 units per month). But production planners are planning
produce 130 units per month
forecast and
in
more even than the
month 16
initial
peak
(-2-) --
of
much more
to
than the
customer orders!! Their
D-3807
1
23
2p
CO
3 rco
130.000
8T
120.000
4^2
<aT
1-2«-
10,000
100.000
\
2
90.000
20.000
10.000
0.0
40.000
30.000
Time
Figure 9: Customer Orders and Production with
Order Monitoring and Control
2
pdi
idi
5.000
2}
4375
2}
3 750
J)
3.125
^
-**
21
2.500
'
0,0
—
'
'
10
I
000
I
<-
—
——— — —— — — — ——r— —1—I—
I
1
'
I
IH
20.000
I
I
I
1
I
I
I
30.000
1
I
n
40,000
Time
Figure 10: Delivery Interval with Order Monitoring and Control
D-3807
plans
24
(-2-)
continue to exceed the forecast up to and beyond month 20,
though the discrepancy gradually diminishes. But how can the planners be
confident they are making the correct decisions? Figure 12 provides the
answer. Schedulers are tracking excess orders
weekly
schedule
So
(-4-).
(-2- in figure
1
1)
and reporting them
month 10 the schedulers report more than 20
to the planners. In
excess orders
(-4-)
the planners increase the monthly production
by 20 units above the forecast.
More Finished Inventor/ and Inventorv Control
The
factory can also regulate delivery interval by holding a
finished inventory (or nearly-finished inventory)
on the inventory
as a
level to production planners.
buffer, to allow the factory to ship
and sending information
The inventory
to
in
fact
it
proposed buffer inventory
production planners
does
is
not.
be used
to
weekly production.
think that the idea of a buffer inventory
the textbook. But
is
promptly regardless of whether
weekly customer orders correspond exactly
You might
comprehensive
The
comes
straight from
crucial difference
that the
is
used as a source of information
when customer orders
to
tell
deviate systematically from
the forecast. Traditional buffers are not used to detect systematic error
in
the forecast, because planners
customer orders and forecast
(relative to forecast) will
month.
If
is
assume
random
--
that the difference
so a glut of orders
be balanced by a corresponding
between
this
month
shortfall next
planners believe their forecasts contain only random (not
D-3807
25
2 apo
SU)f
1
3 psp
4 ss
130.000
120.000
no.ooo
100.000
1
2
3
90 000
20.000
10.000
4J
30.000
40.000
Time
Figure
1 1
:
Forecast and Additions to Planned Orders
with Order Monitoring
1
powii
Ob
?>
powli
and Control
4 uo
450 000
4
35 000
i
400 000
4
25000
; 1
2[
4
350 000
15
000
300.000
5.000
250 000
-5 000
30 000
10.000
40,000
Time
Figure 12: Excess Orders with Order Monitoring and Control
D-3807
26
systematic) error, then they
extreme
But
a buffer inventory
install
and use the forecast alone
glut of orders
to
for production planning.
the fast-moving datacommunications market
in
cover the most
it's
quite
easy
for
forecasts to be systematically low or high for months or even years at a
time.
In
this situation,
indicators
of
pessimistic.
compared
of
movements
whether the
The
the buffer inventory are crucial
in
factory's
inventory level
is
forecasts
are
or
optimistic
monitored frequently (say weekly),
with a standard (perhaps three
months worth
shipments) and the discrepancy reported
average mix
of the
to production planners.
planners should add a fixed proportion of the discrepancy (say one
The
half) to
the base 6-month market forecast.
Holding three months of finished inventory might
proposition. But
an expensive
like
one should remember what the investment
buy - timely information on whether demand
is
seem
(potential
is
customer orders)
exceeding supply (planned production), whether demand
balance with supply, or whether
that
is
is
in
buffer inventory can prevent the
company from suppressing
case the inventory's value
product
line's
A
of sales.
product
line's profit
growth rate
to the
company
(per year)
its
is
own
in
numerical example
margin
for similar
is
will
illustrate.
$1000 (on a $5000
products
is
Let's
unit), that
the
orders.
equal to the
margin multiplied by potential annual growth
profit
volume
exact
lower than supply. For a product-line
growing faster than expected, information from changes
is
In this
it
intended to
in
suppose the
the industry
50 percent per year, and that current
27
D-3807
sales volume
is
1000
finished inventory to
units per year.
stabilize
pace with the industry
trend,
its
company had
the
If
delivery interval,
its
invested
in
orders would keep
and reach 1500 by year-end. Without the
demand, so customer
inventory, the factory inadvertently suppresses
orders remain static at 1000 units per year. The presence of the inventory
creates 500 extra orders, each worth $1000
of $0.5 million
a
in profits, for
total benefit
over the year. The carrying cost of the inventory (assuming
3 months (0.25 years) coverage and a 10 percent interest rate)
$0,125
look
million
more
--
a quarter
and costs
simple calculation above gives the flavor.
real possibility for
If
in
a
to
go together) then
to
self-suppressing
ppm
is
on
demand
investment
inventory
is
possible.
If
g
is
coverage - expressed as a
fraction of
the return on inventory investment
shows a
is
is
it
is
a
quite easy
a general formula
is
in
situations
for
where
the industry growth potential,
the percent profit margin on the product-line, c
Figure 13
but the
be 100, 200 or even 300
percent per year. (As a matter of interest, there
return
--
demand
self-suppressing
a return on finished inventory investment
calculating
real situation
to
a product-line with high growth potential and high
margin (high growth and margin tend
for
only
Of course one would want
of the benefit!
closely at the benefits
is
a year, and
i
is
the inventory
the interest rate, then
(((g*ppm)/(c*i))-1)*100).
simulation of the finished inventory monitoring and
control policy, with a
20 percent increase
should compare figure 13 with figure
5.
As
in
sales
before,
effort.
when
The reader
sales effort
D-3807
28
2p
CO
1
3 rco
1]
2
130.000
11
2
120,000
3J
110 000
2
3J
1)
2
3
100.000
90 000
3)
H
0.0
—
—— — — — —
'
I
I
I
I
I
'
I—'
I
>
'
'
'
— —' ''''
''-I
I
I
•
I
•
1
40 000
30.000
20.000
10.000
Time
Figure 13: Customer Orders and Production with
Finished Inventory Control
1
i)
3
2
pdi
1.500
1.100
i]
4.375
I]
1
eis
250
1.000
3.750
3
4
0.900
41
3.125
1
000
750
-Hj-
0.800
2 500
500
4
4
aqi
5.000
4
3
4
3
idi
0.700
—'— — —
-\
.
00
1-3
-4-^2
'
'
I
i^
1^
—
•~
10.000
«
20.000
1
1-3-
1
1
1
1
1
.
30.000
Time
Figure 14: Delivery Interval with Finished Inventory Control
r—
40.000
29
D-3807
increases, customer orders
per month.
Now
(-1-)
increase correspondingly to 120 units
declining. Meanwhile, production
customer orders
remain constant rather than
tnowever, orders (-1-)
month
in
(-2-)
increases slowly
12. Production (-2-) continues to grow,
about 125 units per month then gradually declines
with orders.
until
The production overshoot enables
until
it
it
equals
peaks
at
balances exactly
the factory to re-build
its
depleted finished inventory back to 3 months worth of shipments.
Figure 14
steady
shows
at 3
that the
months
prepare and ship
unexpected surge
new
policy holds delivery interval
because there
--
is
customer order
all
of orders. In other
(-1-)
rock
adequate inventory on hand
in
to
three months, despite the
words the finished inventory acts as a
buffer to insulate shipments from production.
Figure 15
shows how
inventory also acts as a source of planning information for production
month 5 (when customer orders step-up) inventory
planners.
In
declines,
because shipments exceed production. At the same time
authorized inventory
(-2-)
begins to
rise,
because shipments are now
higher and the factory needs more finished inventory
3
months coverage
inventory
to
(-2-)
of
(-1-)
in
order to maintain
shipments. The difference between authorized
and actual inventory
increase the production plan
-
it's
(-1-) is
the signal that planners use
the analogy of excess orders
in
the
order control policy. Planners add half the inventory discrepancy to the
six-month forecast
in
order to compute planned production six-months
ahead. The simulation shows that the planners' correction
to the forecast
D-3807
30
(-4-) rises to
a peak
of
by month 27 and goes
20
month by month
units per
slightly
the inventory control policy
12, declines to zero
negative between months 28 and 40. With
in
effect,
planners expand the production
schedule more quickly than the forecast alone would suggest
they did with the order control policy. Figure 16 shows what
--
just
as
happening.
is
As
before, the short-term business forecast (-1-) rises slowly from 100
to
120
units per
customer orders
month.
will
month
In
be 110
units per
month
10, planners are preparing
month
16.
They
justify
all
the
to
month
six
in
produce 125
months
units per
time. But
month
(-2-)
in
by
the extra planned production on the basis of the
finished inventory shortfall.
forecast
10, forecasters are predicting that
In fact,
way from month
the shortfall
tells
them
to plan
above
5 to month 25 of the simulation.
extra planned production replenishes and builds finished inventory.
The
D-3807
1
1
2
n
31
2
ai
3
cpi
4
cpoi
32
D-3807
DOCUMENTATION OF THE SALES AND
PRODUCTION SCHEDUUNG MODELS
Policy Structure of the Sales
STELLA Diagram
and Production Scheduling Model
Customer Ordering and Forecasting in the
Base Sales and Production Scheduling Model (saps_base)
of
STELLA Diagram
of Delivery Interval in the
Base Sales and Production Scheduling Model (saps_base)
STELLA Diagram
Scheduling and Expediting in the
Base Sales and Production Scheduling Model (saps_base)
of Production
STELLA
Equations for the
Base Sales and Production Scheduling Model (saps_base)
STELLA Diagram
in
STELLA
Order Monitoring, Production Scheduling and Expediting
the Sales and Production Scheduling Model with
Order Monitoring and Control (saps_ordmon)
of
Equations for the Sales and Production Scheduling Model with
Order Monitoring and Control (saps_ordmon)
STELLA Diagram
Customer Ordering and Forecasting
Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
STELLA Diagram
of
in
and Shipping
Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
of Finished Inventory Control
the
in
the
33
D-3807
DOCUMENTTATION OF THE SALES
AND PRODUCTION SCHEDULING MODELS - CONTINUED
STELLA Diagram
of Delivery Interval in the
Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
STELLA Diagram
Order Monitoring, Production Scheduling and Expediting
in the Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
STELLA
of
Equations for the Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
Description of
(saps_base)
New
to
Structure and Equations to Convert
(saps_ordmon) and
to
(sapsjnvcon)
D-3807
34
.::::: Monufocturinq
P'onninq ::;;::
':::::::::::::::::::::::::::: Horizon •••••
:::::::::::::::
::::::
EXPEDITING
-
;
Production :::::::::::::::: irrrrrrT:
;
:
:
:
:
:
:
Ai-ilite
Figure17: Policy Structure of the Sales and
Production Scheduling Model
R
35
D-3807
Figure 18:
STELLA Diagram
of
Customer Ordering and Forecasting
in
Base Sales and Production Scheduling Model (saps_base)
the
D-3807
36
\r\
rrr\
pVannt A
order i
Figure 19:
STELLA Diagram
of Delivery Interval in the
Base Sales and Production Scheduling Model (saps_base)
D-3807
Figure 20:
37
STELLA Diagram
the
of Production
Scheduling and Expediting
in
Base Sales and Production Scheduling Model (saps_base)
D-3807
38
©
fpo - fpo - p * epo
INIT(fpo) =
stbf»mph
Ob = Ob + CO - s
INIT(ob) =
n
D
O
O
O
edits = grophCrdi)
0.0 -> 0.800
0.500 -> 0.850
300
1.000 -> 1.000
cpdi
1.500 -> 1.150
pdi = pdi
INIT(pd1) = idi
2.000 -> 1.350
stbf = stbf + cstbf
INIT(stbf) = CO
2.500 -> 1.550
3.000 -> 1.825
aporstbf
3.500 -> 2.150
CO = se/ts
4.000 -> 2.500
cpdl = (edi-pd1)/tpdi
4.500 -> 2.900
5.000 -> 3.500
cse =
.2
cstbf = (co-stbf)/taf
O
O
O
O
O
O
d1 =
ob/p
edi = (ob/f po)*mph
fp =
.3
fpoc = ob/fpo
idi =
3
ise =
7500
niph = 6
nts = 75
P = ss*f p+psp*(
O
1
-f p)
powil = fpo*(id1/mph)
psp = fpo/mph
rco = se/nts
O
rdi = pdi/idi
s = p
se =
IF
TIME <5 THEN ise ELSE (ise»(1+cse))
O
ss = ob/pdi
C
taf = 6
O
tpdi =
1
ts = nts*edits
STELLA
Equations for the Base Sales and Production
Scheduling Model (saps_base)
39
D-3807
Figure 21:
STELLA Diagram
of
and Expediting
the Sales and Production Scheduling Model
with
in
Order Monitoring, Production Scheduling
Order Monitoring and Control (saps_ordmon)
D-3807
40
= f po - p + apo
f po
INIT(fpo) =
lj Ob
= ob +
CO - s
INIT(ob) =
Lj
stbf»mph
300
pdl = pdi + cpdi
INIT(pdl) = idi
stbf = stbf + cstbf
INIT(stbf) = CO
apo = stbf +(uo/tcuo)
CO = se/ts
O
cpdi = (edi-pd1)/tpdi
cse =
.2
cstbf = (co-stbf )/taf
O
O
O
O
O
O
cwos
J
ise =
=
ob/p
di =
edi =
(ob/fpo)*mph
fp= 3
fpoc = ob/fpo
idi =
3
7500
O
iwos =
"j
mph
O
nts = 75
C
OS = fpo*fpoc
O
O
p =
=
6
ss*fp+psp*(l-fp)
powii = fpo*(idi/mph)
C-'
powti
C
psp = fpo/mph
O
O
=
fpo*(pd1/mph)*(1-wft)+fpo*(idi/mph)*wft
rco = se/nts
rdi = pdi /idi
rpo = os*wos+powti*(1-wos)
1
s = p
STELLA
Equations for the Sales and Production Scheduling Model
with Order Monitoring
and Control (saps_ordmon)
D-3807
se =
Q
O
O
O
41
IF
TINE <5 THEN ise ELSE (ise*(1*cse))
ss = ob/pdi
taf = 6
tcuo =
tpdi =
1
1
ts = nts*edits
O
O
O
uo = ob-rpo
wft
=
wos
= IF
1
TIME
<
40 THEN wos ELSE
i
(i
wos*( +cwos))
1
edits = graph(rdi)
0.0 -> 0.800
0.500 -> 0.850
1.000 -> 1.000
1.500 -> 1.150
2.000 -> 1.350
2.500 -> 1.550
3.000 -> 1.825
3.500 -> 2.150
4.000 -> 2.500
4.500 -> 2.900
5.000 -> 3.500
STELLA
Equations Continued
--
for the
Sales and Production
Scheduling Model with Order Monitoring and Control (saps_ordmon)
D-3807
Figure 22:
42
STELLA Diagram
in
of
Customer Ordering and Forecasting
the Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
D-3807
Figure 23:
43
STELLA Diagram
in
of Finished Inventory Control
the Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
and Shipping
D-3807
Figure 24:
44
STELLA Diagram
of Delivery Interval in the
Sales and Production Scheduling Model
with Finished Inventory Control (saps-invcon)
45
D-3807
^
D-3807
\
46
aal = asl * cesi
INIT(asl) = si
Hj
fi
=
+ p - si
fi
INIT(fi) = co*8ic
Cj fpo
apo - p
= fpo +
INIT(fpo) =
stbf*mph
ob = Ob + CO - s
INIT(ob) =
Lj
300
pdi = pdi + cpdi
INIT(pdi) = idi
n
slbf = stbf + cstbf
INIT(stbf) = CO
O
O
9i =
asi*aic
aic = 3
apo = stbf+(uo/tcuo)+(cpoi*wic)
O
O
O
O
O
O
aqi = fi/(ss*aic)
casi = (si-asi)/tas
CO = se/ts
cpdi = (edi-pdi)/tpdi
cpi = (ai-fi)/tcip
cpoi = (ai-fi)/tcis
cse =
.2
C
cstbf = (co-stbf)/taf
''_
cwos
di =
=
ob/si
O
edi =
C
fp = -3
;~;
(ob/fpo)*mph*(1-wdi)+di*wdi
fpOC = Ob/fpO
C
idi =
3
O
ise =
7500
iwos
=
O
O
~;
mph
=
1
6
nts = 75
OS = fpo*fpoc
STELLA
Equations
for the
Sales and Production Scheduling Model
with Finished Inventory Control (sapsjnvcon)
D-3807
O
1^
I
p =
47
(ss+cpi)*fp+psp*(1-fp)
powii = fpo*(icli/mph)
powti
O
O
psp
=
fpo*(pdi/mph)*(1-wft)+fpo*{id1/mph)*wft
= f po/mph
rco = se/nts
rdi = pdi/idi
O
rpo = os*wos+powti*(1-wos)
'.0
s = si
se =
si =
TIME <5 THEN ise ELSE (ise*(l+cse))
IF
ss*eis
ss = ob/pdl
C
taf = 6
C
tas= 12
O
tcip = 2
tcis = 2
tcuo =
O
O
I
J
tpdi =
1
1
ts = nts*edits
uo = ob-rpo
C' wdl =
1
wft
=
wic
=
wos
= IF
1
TIME
<
40 THEN iwos ELSE (1wos*(1*cwos))
edits = graph(rdi)
0.0 -> 0.800
0.500 -> 0.850
000->
!
eis = graph(aqi)
0.0 ->
0.0
0.200 -> 0.950
1.000
0.400 -> 0.980
1.500 -> 1.150
0.600 -> 1.000
2.000 -> 1.350
0.800 -> 1.000
2.500 -> 1.550
1.000 -> 1.000
3.000 -> 1.825
1.200 -> 1.000
3.500 -> 2.150
1.400 -> 1.000
4.000 -> 2.500
1.600 -> 1.000
4 500
-> 2.900
1.800 -> 1.000
5.000 -> 3.500
2.000 -> 1.000
1
STELLA
Equations Continued
--
for the
Sales and Production
Scheduling Model with Finished Inventory Control (sapsjnvcon)
48
D-3807
DEFINITION
OF VARIABLE NAMES
ai
casi
authorized inventory (systems)
authorized inventory coverage (months)
additions to planned orders (systems/month)
average shipments from inventory (systems/month)
adequacy of inventory (dimensionless)
change in average shipments from inventory
CO
(systems/month/month)
customer orders (orders
aic
apo
asi
aqi
cpdi
cpi
cpoi
cse
cstbf
systems/month)
change in perceived delivery interval (months/month)
correction to production from inventory (systems/month)
correction to planned orders from inventory (systems/month)
change
change
in
in
for
sales effort (dimensionless)
short-term business forecast
(systems/month/month)
cwos
change
di
delivery interval (months)
edi
estimated delivery interval (months)
effect of delivery interval on time per sale (dimensionless)
effect of inventory on shipments (dimensionless)
finished inventory (systems)
flexibility of production (dimensionless)
firm planned orders (systems planned)
fraction of planned orders committed (dimensionless)
industry delivery interval (months)
initial sales effort (hours/month)
initial weight for orders scheduled (dimensionless)
manufacturing planning horizon (months)
normal time per sale (hours/system)
orders booked (orders for systems)
orders scheduled (systems planned)
production (systems/month)
published delivery interval (months)
planned orders within industry interval (systems planned)
edits
eis
fi
fp
fpo
fpoc
id!
ise
iwos
mph
nts
ob
OS
p
pdi
powii
in
weight
for
orders scheduled (dimensionless)
49
D-3807
DEFINITION
OF VARIABLE NAMES - CONTINUED
powti
psp
rco
planned orders within target interval (systems planned)
production suggested by plan (systems/month)
reference customer orders (orders for systems/month)
rdi
relative delivery interval (dimensionless)
rpo
reference planned orders (systems planned)
s
se
shipments (systems/month)
si
ss
stbf
tat
tas
tcip
tcis
tcuo
tpdi
ts
uo
wdi
wft
wic
wos
sales effort (hours/month)
shipments from inventory (systems/month)
scheduled shipments (systems/month)
short-term business forecast (systems/month)
time to adjust forecast (months)
time to average shipments (months)
time to correct inventory for production (months)
time to correct inventory for schedule (months)
time to correct unexpected orders (months)
time to publish delivery interval (months)
time per sale (hours/system)
unexpected orders (orders for systems)
weight for delivery interval (dimensionless)
weight for fixed target (dimensionless)
weight for inventory correction (dimensionless)
weight for orders scheduled (dimensionless)
D-3807
50
DESCRIPTION OF PARAMETER AND STRUCTURAL CHANGES FOR SIMULATION
SCENARIOS
Base Run (Model saps_base)
The base
run
is
described on pages 15 through 19 of the report. The base
run uses the model
no feedback
is
to
saps_base
to production
--
a version
of the
model
planning from orders booked, and
no finished inventory. The production plan (more
planned orders)
assumed
is
smoothing constant (time
of
which there
in
is
which there
specifically, additions
therefore equal to the forecast.
be an exponential smoothing
to
in
The forecast
is
customer orders, with a
to adjust the forecast taf) of
6 months.
Monitoring and Control of Excess Orders (model saps_ordmon)
The model saps_ordmon
is
the
same as
the base model, but with
new
equations to represent a policy of monitoring and control of excess orders.
Simulations of the model are described on pages 20 through 24 of the
report.
a).
The new equations are described below.
Additions to Planned Orders
apo =
tcuo =
stbf
1
+ uo/tcuo
1
2
where:
apo
additions to planned orders (systems planned/month)
stbf
uo
short-term business forecast (systems/month)
unexpected orders (orders for systems)
tcuo
time to correct unexpected orders (months)
D-3807
In
51
the base model, the additions to planned orders (apo) are equal to
short-term business forecast
new
(stbf).
In
other words,
when planners add
production orders to the production schedule, they add the
orders called for by the forecast
(stbf).
But,
when
the
number
of
they adopt the order
monitoring and control policy they also add a proportion of the unexpected
orders (uo) on top of the forecast.
more
specifically,
when
it
is
b).
the forecast
is
innaccurate, or
biased downward, unexpected orders
accummulate which are factored
equation
When
into the
production plan as
shown
in
1
Unexpected Orders
But what are unexpected orders and
how do schedulers recognize them?
Equations 3 through 9 show how.
uo = ob - rpo
rpo = os*wos + powti*(1-wos)
OS = fpo*fpoc
fpoc = ob/fpo
wos =
where:
D-3807
52
wos
powti
pdi
for
industry delivery interval (months)
idi
mph
wft
The equations
and weight
orders scheduled (dimensionless)
planned orders within target interval (systems)
published delivery interval (months)
weight
manufacturing planning horizon (months)
weight for fixed target (dimensionless)
contain two switches
for fixed
--
target (wft) which allow
assumptions about the effectiveness
The simulations described
policy.
wos =
and
effective
wft =
1
weight for orders scheduled (wos)
--
in
of the
one
to
make
different
order monitoring and control
the report were obtained by setting
a combination
of
order monitoring and control.
switches that results
in
very
With this combination, the
equation for reference planned orders (rpo) reduces
to:
rpo = fpo*(idi/mph)
So,
we
are saying
in
equation 3 that schedulers recognize unexpected
orders (uo) by comparing orders booked (ob) with reference planned orders
(rpo),
and
scheduled
that they take as their reference point only the planned orders
for
production within the industry's delivery interval (the
quantity fpo*(idi/mph).
If
the parameter, weight for orders scheduled (wos),
model saps_ordmon becomes equivalent
to the
is
set to
1,
then the
base model. Under
this
condition, the equation for reference planned orders (rpo) reduces to:
rpo = OS
But,
as equations 5 and 6 show, orders scheduled (os) are those planned
orders that have been assigned to customer orders
orders booked (ob)!! So,
in
equation
3,
--
in
other words,
unexpected orders are always zero,
D-3807
and
in
53
equation
1
additions to planned orders (apo) are equal to the
short-term business forecast
(stbf)
--
the
same assumption used
base model. (One might ask, why take the trouble
reference planned orders (rpo)
orders (rpo)
is
zero.
if
The answer
to write
in
the
equations
for
the numerical value of reference planned
is
that the equations
show
which schedulers and planners recognize excess orders
--
the process by
the information
they use to assess whether customers have ordered more than expected.
turns out that
in
customer orders
difficult to
a system where schedulers have the freedom
to
any open production
slot (a slot-planning
recognize excess orders because,
if
to
It
assign
system)
it's
the planning horizon
is
long enough, schedulers can always find open production slots).
Finished Inventory Control (model sapsjnvcon)
The model sapsjnvcon
added
to
is
same as saps_ordmon,
but with equations
represent finished inventory, production for inventory, shipping
from inventory
(in
saps_ordmon shipping
finished inventory control.
a).
the
is
equal
to production)
The new equations are described below.
Finished Inventory and Shipping
ufi =
+ p - si
INIT(fi) = co*aic
si = ss*eis
eis = graph(aqi)
aqi = fi/(ss*aic)
fi
1
2
3
4
5
and
D-3807
54
where:
finished inventory (systems)
fi
production (systems/month)
shipments from inventory (systems/month)
customer orders (orders for systems/month)
authorized inventory coverage (months)
p
si
CO
aic
eis
scheduled shipments (systems/month)
effect of inventory on shipments (dimensionless)
aqi
adequacy
ss
Equation
(p)
1
of inventory (dimensionless)
states that finished inventory
(fi)
is
and reduced by shipments from inventory
shipments are no longer
identical
as they were
increased by production
(si).
in
So production and
the base model and
in
saps_ordmon, but can move (somewhat) independently because they are
separated by a
level of inventory.
Equations 3 through 5 describe shipping.
inventory then shipments from inventory
schedule
(ss).
The
(si)
there
is
adequate finished
are equal to the shipping
factory can ship according to schedule
adequate product available
shipments
If
(eis) is neutral
--
in
because there
is
other words, the effect of inventory on
(takes a numerical value of
1).
The key
to the
success
of the finished inventory control policy is for the factory to hold
enough
finished inventory that stockouts of high-volume items never
occur.
By preventing stockouts, the
intervals
and so avoid the
factory can maintain constant delivery
possibility
of
inadvertently
suppressing
customer orders. The following parameter values represent a no-stockout
inventory
in
the model sapsjnvcon:
55
D-3807
authorized inventory coverage aic = 3 months
on shipments eis = graph(aqi) such that
effect of inventory
when adequacy
It
is
of inventory
aqi =
eis =
aqi = 0.2
eis = .95
aqi = 0.4
eis = .98
aqi = 0.6
eis
=1.0
aqi = 0.8
eis
aqi = 1.0
eis
aqi = 1.2
eis
aqi = 1.4
eis
aqi = 1.6
eis
aqi = 1.8
eis
=
=
=
=
=
=
aqi = 2.0
eis
=1.0
1.0
1.0
1.0
1.0
1.0
1.0
important that the effect of inventory on shipments
the value
1,
or very close to
inventory,
because then the
schedule.
If
shipments
you'll
see
that
is
always able
when adequacy
(eis) is
also
1.
When
the factory's inventory
to ship
one
fifth
start to
of inventory
--
on
of inventory
is
1
--
when
the
is
the
then the effect of inventory on
--
adequacy
of inventory (aqi) falls
less than authorized
reaches the value
of the authorized
acording to
for the effect of inventory
inventory on shipments (eis) stays very close to the value
adequacy
at
exactly equal to the authorized inventory,
three months of shipments
is
shipments
when
factory
is
remains
over a wide range of values of finished
you study the graph function
factory's finished inventory
which
1,
(eis)
does the
.2 --
when
--
1
.
decline rapidly, signifying that the factory
the effect of
Only when the
the factory's
effect of inventory
is
--
is
only
on shipments
out-of-stock on
D-3807
some
56
product lines and therefore unable to ship according to schedule.
The reader should be aware
that
on shipments
of the effect of inventory
authorized inventory coverage
factory produces
many
and holds
in
different products then
inventory
(in
one cannot
(eis).
arbitrarily specify
The shape depends on
and on the
(aic)
diversity of products
finished inventory.
it
If
of
shipments)
should plan to hold a
to
be sure
the
the
the factory produces
lot
other words, the authorized inventory coverage
say 3 or 4 months
the shape
of finished
be
will
high,
of avoiding stock-outs,
and
therefore to be sure that the effect of inventory on shipments remains
close to the value
product
So
On
the other hand,
if
the factory produces only one
a stock-out can occur only when finished inventory
line,
the effect of inventory on shipments (eis)
even
or
1.
if
when
of
about the shape
it
is
of the effect of inventory
changes depending on the
if
one were
coverage reduced
important to think carefully
on shipments
factory's authorized inventory
to rerun
to only
sapsjnvcon
one month, instead
it
would be essential
to
make
(eis)
coverage
- how
(aic)
it
and
of product lines.
with authorized inventory
of three,
assumption that the factory produces several, say
then
only one month
is
on the assumptions one makes about the factory's number
lines)
1
shipments.
using the model sapsjnvcon,
For example,
zero.
remain at the value
the factory's authorized inventory coverage (aic)
one week
So,
will
is
(and with the
10, different product
the effect of inventory on
57
D-3807
shipments
would
(eis)
find
inventory
slope more gradually upward from the (0,0) point.
the
that,
more gradual the
policy
control
would be
in
because the factory would stock-out more
b).
slope, the
regulating
One
less effective the
delivery
interval
--
easily.
Finished Inventory Monitoring and Control
To use the
factory's
finished inventory effectively for production planning, the
warehouse must monitor
the
between finished
difference
inventory and authorized inventory and report the difference to schedulers
and planners. The schedulers and planners use the information
both current production and the production plan.
The equations
the monitoring and control of finished inventory are
apo =
stbf + (uo/tcuo) + (cpoi*wic)
cpoi =
(ai
-
fi)/tcis
p = (ss + cpi)*fp + psp*(1-fp)
cpi = (ai - fi)/tcip
to adjust
to represent
shown below:
6
7
8
9
where:
apo
additions to planned orders (systems/month)
stbf
short-term business forecast (systems/month)
unexpected orders (orders for systems)
time to correct unexpected orders (months)
correction to planned orders from inventory
uo
tcuo
cpoi
(systems/month)
ai
fi
tcis
p
ss
authorized inventory (systems)
finished inventory (systems)
time to correct inventory for scheduling (months)
production (systems/month)
scheduled shipments (systems/month)
D-3807
58
correction to production from inventory
cpi
(systems/month)
production (dimensionless)
production suggested by plan (systems/month)
time to correct inventory for production (months)
flexibility of
fp
psp
tcip
Equation 6 states that when planners
make
they add to the short-term busines forecast
orders from finished inventory (cpoi). They
there
is
(fi)
as shown
short-term business forecast
-
(stbf)
make a
but
in
the simulations
in
(stbf)
shown
equation
7.
scheduled
in
correction
planned
whenever
(ai)
(They also add
to the
(rpo) are equal to orders
(os)).
make a
correction to production
based on the reported difference between authorized inventory
and finished inventory
authorized
order to
and
the report, unexpected orders are
Equation 8 and 9 state that schedulers
in
to
a correction for unexpected orders
always zero because reference planned orders
(cpi)
a correction
a reported difference between authorized inventory
finished inventory
(uo)
additions to planned orders
(ai)
(fi).
If
finished
inventory
(fi)
is
(ai)
lower than
they add a portion of the difference to scheduled shipments
consume
the schedule of planned orders
re-build finished inventory.
more
quickly
and so
59
D-3807
BACKGROUND READINGS IN SYSTEM DYNAMICS
Forrester J.W. Industrial
Dynamics
,
M.I.T. Press,
Cambridge, MA, 1961.
Forrester J.W. 'Market Growth as Influenced by Capital Investment', Sloan
Management Review, Vol 9, No 2, pp 83-102, Winter 1968.
Lyneis J.M. Corporate Planning and Policy Design:
Approach, M.I.T. Press, Cambridge, MA, 1980.
A System Dynamics
Model Behavior' System Dynamics Group
Working Paper D-3323, Sloan School of Management, M.I.T., Cambridge, MA,
Mass
N.J. 'Diagnosing Surprise
October 1981.
Morecroft J.D.W. 'A Behavioral Model of Sales Planning and Control in a
Datacommunications Company', Sloan School of Management Working Paper
WP-1761-86, and System Dynamics Group Working Paper D-3806, M.I.T.,
Cambridge, MA, March 1986.
Morecroft J.D.W. 'The Feedback View of Business Policy and Strategy'
System Dynamics Review, Vol 1 No 1 pp 4-19, Summer 1985.
,
,
Morecroft J.D.W. 'Strategy Support Models', Strategic
Vol 5, No 3, pp21 5-229, August 1984.
Pugh
A.L.
DYNAMO User's Manual, 6th
Ed., M.I.T. Press,
Management Journal,
Cambridge, MA, 1983.
Richardson G.P. and Pugh A.L. Introduction to System Dynamics Modeling
with DYNAMO, M.I.T. Press, Cambridge, MA, 1981.
Richmond B.M. A User's Guide to STELLA - 2nd printing, High Performance
Systems, RR1, Box 37, Lyme, N.H. 03768, December 1 985.
Roberts E.B. Managerial Applications of System Dynamics, M.I.T. Press,
Cambridge, MA, 1978.
8830 019
vA.
Date Du&^r
Lib-26-67
Mir
I
3
lllll
^IQflD
III
I
III
mwAKits
I
Mill
DDM
lllll
II
lljlll
DflM
III
II
3DM