1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
System Dynamics Model of the Standards Development Process
Henry Neimeier
The MITRE Corporation
7525 Colshire Drive
McLean, Virginia, 22102, USA
Abstract
A overall dynamic model of the dard: lop process was developed to d savings
from dards i in the Defense Information System. The model
will aid in allocating standards development resources. Different funding and personnel strategies
are quantitatively compared.
Industry, page 68
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
System Dynamics Model of the Standards Development Process
Overview
An overall dynamic model of the standards development process was developed to document
savings btainable from dard: in the Defense Information System
(DIS). DIS is defined as the aggregation of all DOD i ion systems, including sensors, data
entry devices, displays, icati ‘ks, and P The model provides
a 20-year view of the ication of dards to DIS Quantitative measures of
performance are calculated and dards resource allocation is optimized.
Introduction
The standards development process may be described in terms of many interacting factors with
nonlinear feedbacks. Sheer intuitive judgement is unreliable about how the total process will
change with time, even when there is good knowledge of the individual parts of the system.
Standards lifetimes can be 20 years or more. Similarly the time to identify potential standards
projects and complete the standard can take years. Thus, a long-range perspective is required.
The system dynamics model encodes the underlying standards process in an easily understood
structure. Quantitative measures of performance are provided to defend standards budgets,
document savings, and guide resource allocation. These include standards coverage, Proportion of
potential savings achieved. total dollars saved, and dards costs. Al
strategies are evaluated and optimal p i ies are developed. he’ effect of standard
budget variation on standards process performance is itati i itive policy
parameters are identified including: labor productivity, relative workload, relative wage rate, and
training. The model can be extended to monitor existing dards activities and d:
estimates.
Model Implementation
The model was implemented in the Stella systems dynamics simulation language on an Apple
Macintosh computer. Stella is easy to learn, provides animated graphics, interfaces to
spreadsheets and object oriented graphics programs. Stella also provides a direct interface to
Apple Hypercard. This was used to develop an animated teaching game from the model . Figure
1 is a diagram of the complete model. Conserved flows such as projects, dollars, or people are
shown with double lines (pipes). Information flows are shown with single lines. Information flows
are instantaneous while conserved flows take time to change. Levels or stocks are shown as
rectangles. The shading indicates the relative amount of stock at that point in time. Valves
(circles with a T) indicate rates of flow into or out of a level. The position of the arrow within
the circle indicates the relative rate of conserved flow. Auxiliary values used in calculating rates
or measures are shown with circles.
The model is divided into three sections: projects, savings, and personnel. These are
named by the conserved flow represented. The projects and savings sections each have five
levels. Project levels are: potential. identified, standards, obsolete, and lost. The potential
projects level represents the number of potential DIS standards projects. This value is estimated
based on the success of the identification effort. New projects enter this level at the technology
growth rate (TGR). The identification task consists of a description of the standard project,
estimate of annual standard savings, standard life, and staff months to write the standard.
Projects are identified at the identification rate (IDR) which is dependent on the number of
personnel assigned (PER_ID) and labor productivity (ID_P_PER). Identified projects join the
Industry, page 69
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
identified level (ID_PROJ). They are ranked in savings times standard life divided by writing cost
order. The most cost effective projects are chosen for writing and Join the standards level
(S1D§): at the writing rate (WR). Writing rate is d dent on igned and labor
ity. C d dards join the obsolete standards level (OBS_STDS) at the
obsolescence rate (OBS | S). Projects are lost (join the PROJ_LOST level) while awaiting
identification and writing based on the standard project life. The model calculates identification
and standard: ge P
The savings section just below the projects section has similar levels except now the
conserved flow is dollar savings. Beta distributions of standard project life and cost were fit based
on initial efforts at identifying projects and writing standards. Since the most cost effective
identified projects are first selected for writing, cost effectiveness would decrease through time if
there was no growth in technology. Initially most p | are all d to the i fi
task, so there is a wide range of identified projects ‘to choose from. Later more personnel are
allocated to the standards writing and approval process. Some personnel remain in the
identification task to handle the technological growth rate. Once a standard is written and
accepted, annual savings are available for the rest of its life. The savings remaining (SAV_REM)
level includes all these savings. As time progresses a proportion of these savings (SAV_P_YR
rate) are taken and join the savings taken (SAV_TAKEN) level. Note that if all standards efforts
ceased, savings in the savings remaining level will still be taken. Cost effectiveness is the savings
taken per year divided by the annual standards budget. Savings proportion is the total savings
taken and remaining divided by the potential savings (savings in all the savings levels). With
infinite personnel or labor productivity, all potential projects would instantaneously be identified
and written and no projects would be lost while waiting. This would result in a unity savings
proportion.
The bottom personnel and budget section models personnel retention, training, labor ©
productivity and budget variation. Labor ion is key to effecti dards op In
international standards bodies it takes years to be effective in dard: i i
gain positions of influence, and obtain standard approvals.
Bt
Industry. page 70
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
COV_SAV
TRAIN.PROP TRAIN.EFF
NEW cosT_cuM u
HIRE_W_MUL COST_P_YR H
LIFE_STD
cvar_pup VAR_EFF ENVIRON REL_WAGE
Figure 1. Complete Model
Industry, page 71
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Results
Figures 2 through 7 show the results of the baseline standards model run over a 20 year period.
Figure 2 shows the project levels. Initially there is a fast decline in potential projects as they are
ified. Later pleted dards increase at a decreasing rate. Both obsolete standards and
lost while waiting levels accumulate values from the beginning of the simulation. Thus they
constantly increase.
-*- Potential
250 5. Identified
~. Standards
obec! =
200 LE standards = 3
“_Lost while waiting a
a)
He 150
® ao ee eee
° ie
100
OU
pe BE =
on ' jy
0123 45 67 8 9 101112 131415 1617 18 1920
Year
Figure 2. Baseline Standards Projects
Figure 3 shows savings in millions of dollars. The first three levels reach stability around year
eight. Note since savings taken and lost accumulate values from the beginning of the simulation,
they increase throughout the run.
-4- Potential
250 a Identified
-.- Standards
fe)
wee
200 < standards _
*_Lost while waiting ao
ra 150+ a
o we = ee ee
° x and
© “
p- an
—
0123 45 67 8 9 101112 131415 1617 18 1920
Year
Figure 3. Baseline Savings
Industry, page 72
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Figure 4 shows the optimal allocation of p 1 between the identi ion and writing tasks
for the baseline values. Initially everyone is assigned to identification. There is a speedy
reallocation to writing which stabilizes in year 4. Identification personnel increase after year 4 in
an attempt to keep up with the technology growth rate. In the base case technology is growing
at 10 percent per year while the standards budget is growing at 5 percent per year.
35
od
<a a
yp ¢
—
ne
-"- Identify P 1
ca i! +2 Writing Personnel
Personnel
\ ee
5
0 rs ett ee om)
0123 456 7 8 9 1011 1213141516 1718 1920
Year
Figure 4. Personnel Allocation
Figure 5 shows the performance proportions. Identification coverage reaches 90 percent by year
4. Standards coverage ii at a di ing rate through the run, hing 65 percent
coverage in year 20. Savings proportion peaks in year 5 at 58 percent and slowly decreases
thereafter. This is because less cost effective projects must be chosen as coverage increases. The
results are highly dependent on the parameter values chosen, so a sensitivity analysis around the
baseline values was performed.
0.9 a————_——E—EAEE
0.8
5 os of — = ae
8 i a
tc 05 +
So. WZ —
9 hi =
2.
0.3 a -"- Savings proportion —
0.2 | oe > Identification coverage |__
. eS -*- Standards coverage
FA ee OS
01234 567 8 9 10111213 1415161718 1920
Year
Figure 5. Performance Proportions
Industry, page 73
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
Figures 6 and 7 present the results of this analysis. The center bar is the base case value. The bar
to the left gives the result when this parameter is set to the low level and all other parameters are
at their baseline values. Similarly the bar above is when the parameter is set to the high level.
The high and low levels are indicated in the parameter label. The baseline level is midway
between these values.
The first four parameters are environmentally set and are thus not controllable. They
are maximum project savings, maximum standard life, technology growth rate, and initial
potential standards projects. The project savings were beta distributed between 1 million dollars
per year and the specified maximum value. The minimum maximum range was set to 6 standard
deviations and the mean was set midway between the minimum and maximum. Standard life was
similarly beta distributed between 5 years and the specified maximum life. To qualify as a project,
savings must exceed one million dollars per year and standard lifetime must exceed 5 years.
The last four parameters are controllable to some degree. They include annual standards
budget, annual standards budget growth rate, identifications per person, and relative government
wage rate. Initially the standards writing rate was set to a tenth of the identification rate (takes
ten times the personnel to write the standard as to identify it).
Figure 6 shows total savings (both taken and remaining) by year 20 in billions of dollars. The
wide range in values is due to the initial uncertainty in parameter values. The range will be
reduced as more estimates are received. Note that there are lower savings for the high
technological growth rate. At 15 percent technological growth and 5 percent standards growth
there are insufficient personnel to keep up with the identification task. An increase in initial
potential projects from the 200 baseline to the 300 high level results in very little increased
HHH | | 4] |
rer, ora
Beeee @
I : fl
oe,
savings for the same reason.
| u
Sav. a a Hi Tech. al Pot.Proj Budget Bud. a ID.P.Per Rel.Wage
10-30$M 10-30Yr 5%-15% 100-300 2-6$M 00-10% 1-9 .65-1.05
Figure 6. Total Savings By Year 20
a
Gj
o
i=]
nN
6
Billion Dollars
a a
a 3
—
Figure 7 shows standards coverage in year 20. Note that the annual savings per project
distribution has no effect on coverage. Increases in standard lifetime, technological growth rate,
and initial potential projects all reduce dard Stand: ge does not include
lost or obsolete projects. Longer standard life leads to more projects in the potential or identified
levels since they take longer to become obsolete or lost.
Industry, page 74
1994 INTERNATIONAL SYSTEM DYNAMICS CONFERENCE
T
Mi Low
. yoas
ao
e a
@
>
°
>
0
Sav.Max rae ae Pot.Proj Budget Bud.Gth ID.P.PerRel.Wage
10-30$M 10-30Yr 5%-15% 100-300 2-65M 0-10% 1-9 .65-1.05
Figure 7. Standards Coverage In Year 20
Reference
Neimeier,A . 1989. System Dynamics Model of the Standards Development Process, MTR-
89W00177, The MITRE Corporation, McLean, Virginia.
Richardson, G. ,Pugh,A. 1981. Introduction to Systems Dynamics Modeling with Dynamo, The
MIT Press, Cambridge, Massachussets.
Richmond, B. et AI.1987,Bussiness User's Guide to Stella, High Performance Systems, Inc.,
Lyme, New Hampshire.
Richmond, B. et Al., 1989.User's Guide to Stella Stack, High Performance Systems, Inc., Lyme,
New Hampshire.
Industry, page 75
