Professional Documents
Culture Documents
Acquisition Programs
Jim McCurley
Senior Member of the Technical Staff
Software Engineering Institute
Carnegie Mellon University
October 2014
_____________________
40% of accumulated
cost overruns
Sources Cost and time overruns for Major Defense Acquisition Programs, Joachim Hofbauer Gregory Sanders Jesse Ellman David Morrow,
Center for Strategic and International Studies, April 2011
QUELCE October 2014
2014 Carnegie Mellon University
Challenge:
Can this knowledge (awareness) be exploited in developing better estimates
and plans for the product development work?
Examples:
Technology introduction: Introduction of 1-10mm wavelength radio frequency
allows much higher volume traffic but development of a modem is more
complicated.
Change in scope: Antenna that fits a 747 will not fit in fighter jet.
Change in interfacing external system
Change in sponsorship: Mission capability of proposed system fits a need for
a new user community provided that certain performance can be improved.
10
Recommendations
11
Challenges to Implementation
Calibrating subject matter experts for probability judgment
Subject matter experts are (almost always) over-confident.
12
...
Operational Capability
Trade-offs
Mission / CONOPS
Capability Based Analysis
...
Technology Development
Strategy
Production Quantity
Acquisition Mgt
Scope definition/responsibility
Contract Award
Probabilistic
Modeling (BBN)
& Monte Carlo
Simulation
Cost Estimates
analogy
parametric
engineering
CERs
Program Execution
Scenarios Evaluated
13
2
2
1
2
2
1
Acquisition Management
Funding Schedule
Scope Definition
Functional Measures
Systems Design
Interoperability
Advocacy Change
Capability Definition
Queries of Historical
MDAP Experience and
Context
3
3
3
3
2
0
3
2
2
1
1
3
2
1
1
1
2
2
1
1
2
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Drivers
Mission / CONOPS
Change in Strategic Vision
Capability Definition
Advocacy Change
Closing Technical Gaps (CBA)
Building Technical Capability & Capacity (CBA)
nteroperability
Systems Design
Mission / CONOPS
Effects
4. Apply Uncertainty
to Cost Formula
Inputs for Basis and
Scenarios
Causes
QUELCE Change
Repository
DAES,
etc.
Scope
3. Assign
Conditional
Probabilities
Interdependency
1. Elicit
Change
Drivers and
Alternatives
2
2
1
2
2
1
1
1
1
XL
VL
6.20
5.07
7.07
5.48
7.80
4.96
4.05
5.65
4.38
6.24
0.49
0.60
2.12
1.59
1.43
1.62
1.33
1.30
1.43
0.83
0.95
0.87
1.26
1.12
1.10
1.14
PREC
FLEX
RESL
TEAM
PMAT
RCPX
RUSE
PDIF
PERS
PREX
FCIL
SCED
Stable
Users added
Additional
(foreign)
customer
defined
New condition
New mission
Capability
Definition
Stable
Addition
Funding
Schedule
Established
Stable
Advocacy
Change
Closing
Technical
Gaps (CBA)
Legend:
Selected Trade
studies are
sufficient
Technology is
too expensive
Selected solution
cannot achieve
desired outcome
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
Complexity
Reduction
2.72
1.24
2.61
0.50
0.62
0.62
Technology not
New technology not
performing as
testing well
expected
~~~~
<X>
<X>
<X>
<X>
<X>
X
X
X
<X>
<X>
<X>
<X>
3. BBN Model
~~~~
Product Project
Alternative States
Additional
Production
Scope Reduction
deliverable (e.g.
downsized
(funding reduction)
training & manuals)
Program
New echelon
becomes Joint
Trade-offs
Subtraction
Variance
[performance vs
affordaility, etc.]
Obligated vs.
FFRDC ceiling Funding change for Funding spread
allocated funds
issue
end of year
out
shifted
Advocate
Service owner
Senator did not Change in senior
requires change
different than
get re-elected
pentagon staff
in mission
CONOPS users
scope
Scope
Definition
Mission /
CONOPS
XH
0.00
0.00
0.00
0.00
0.00
0
1
N
H
VH
Scale Factors
3.72 2.48 1.24
3.04 2.03 1.01
4.24 2.83 1.41
3.29 2.19 1.10
4.68 3.12 1.56
Effort Multipliers
1.00 1.33 1.91
1.00 1.07 1.15
1.00 1.29 1.81
1.00 0.83 0.63
1.00 0.87 0.74
1.00 0.87 0.73
1.00 1.00 1.00
~~~~
Modeling
Uncertainty
QUELCE October 2014
2014 Carnegie Mellon University
14
Broad and diverse literature on eliciting expert judgment exists, especially in Psychology
and Statistics (see Selected References).
Estimates for budgeting, including lifecycle costs, are initially made when the system is
only a concept.
The minimum life cycle for major systems is several years and often spans decades.
Contextual uncertainties, particularly assumptions which can change over time, are not
captured by cost models.
Methods exist to help sharpen expert judgment, which also allow us to generate data to
incorporate the uncertainty associated with that judgment.
[expert] judgments of probability, however elicited, are just that judgments ... made in
response to the facilitators questions, not pre-formed quantifications of pre-analyzed
beliefs. - Uncertain Judgements: Eliciting Expert Probabilities, Anthony OHagan, et.al.,
Wiley, 2006.
QUELCE October 2014
2014 Carnegie Mellon University
15
16
16
Step 2: Iterate
through a series
of domain
specific tests
Step 3: Feedback on
test performance
Step 1: Virtual
training using
reference
points
Outcome: Expert
renders calibrated
estimate of size
DoD Domain-Specific
reference points
1)
2)
3)
Un-Calibrated
Calibrated
Calibrated = more
realistic size and
wider range to
reflect true expert
uncertainty
Estimate of SW Size
Used with permission from Douglas Hubbard Copyright HDR 2008 dwhubbard@hubbardresearch.com
QUELCE October 2014
2014 Carnegie Mellon University
17
18
Generic Tests
Domain Specific
Tests
19
20
2.Reduce complexity
of Cause and Effect
relationships via
matrix techniques
Users added
As defined
New condition
Capability
Definition
Stable
Addition
Funding
Schedule
Established
Stable
Selected Trade
studies are
sufficient
Scope
Definition
Mission /
CONOPS
Advocacy
Change
Closing
Technical
Gaps (CBA)
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
3. Assign
Conditional
Probabilities to
BBN Model
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Alternative States
Additional
Production
Scope Reduction
deliverable (e.g.
downsized
(funding reduction)
training & manuals)
Program
New mission
New echelon
becomes Joint
Trade-offs
Subtraction
Variance
[performance vs
affordaility, etc.]
Obligated vs.
FFRDC ceiling Funding change for Funding spread
allocated funds
issue
end of year
out
shifted
Advocate
Service owner
Senator did not Change in senior
requires change
different than
get re-elected
pentagon staff
in mission
CONOPS users
scope
Additional
(foreign)
customer
Technology is
too expensive
Selected solution
cannot achieve
desired outcome
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
Technology not
New technology not
performing as
testing well
expected
~~~~
~~~~
21
Example:
Expert Data
From:
Obligations & Expenditures:
An Investigation into the Factors that Affect
OSD Goals, DoD Approved Survey:
Report Control Symbol DD-AT&L(OT)2513
Rob Tremaine, Donna Seligman, Shandy
Arwood, John Higbee
Briefing Presented to:
Honorable Katrina McFarland
05 Feb 2013
22
23
24
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
3. Assign
Conditional
Probabilities to
BBN Model
2. Reduce complexity
of Cause and Effect
relationships via
matrix techniques
5. Monte Carlo
Simulation to
Compute Cost
Distribution
2
2
1
2
1
2
2
1
1
1
1
1
1
1
3
2
1
1
1
1
1
0
1
2
1
2
2
2
1
1
2
2
1
2
1
1
2
1
1
2
2
1
2
2
2
1
1
1
1
2
1
1
1
1
2
1
1
1
1
2
2
2
1
1
1
1
3
3
1
2
2
1
1
1
1
2
1
1
1
1
2
2
2
6
0
4
1
1
1
9
4
5
4
12
4
8
1
7
2
7
0
13
3
4
1
10
3
15
2
18
2
7
3
7
1
8
0
8
1
14
0
17
0
17
0
15
0
12
0
9
0
10
0
13
0
11
0
20
0
19
0
5
0
5
0
17
0
6
29
16
6
34
27
29
21
33
16
5
10
5
19
12
14
6
5
6
10
7
7
4
0
0
2
2
0
0
0
2
0
0
0
Total
1
1
Product Challenge
2
2
Project Challenge
2
2
2
1
2
Size
1
1
1
3
2
Contractor Performance
1
1
1
1
1
2
0
1
1
1
1
1
2
1
Cost Estimate
1
2
1
2
1
3
1
1
1
1
Data Ownership
2
2
2
1
2
1
Production Quantity
1
2
Contract Award
2
1
2
2
Sustainment Issues
2
2
2
1
1
1
2
PO Process Performance
1
1
1
Information sharing
2
0
2
2
3
2
2
1
0
1
Standards/Certifications
2
2
1
0
1
3
3
Scope Responsibility
Acquisition Management
Funding Schedule
Scope Definition
Functional Measures
Interdependency
Systems Design
Interoperability
Advocacy Change
3
3
Capability Definition
Mission / CONOPS
Change in Strategic Vision
Capability Definition
Advocacy Change
Closing Technical Gaps (CBA)
Building Technical Capability & Capacity (CBA)
Interoperability
Systems Design
Interdependency
Functional Measures
Scope Definition
Functional Solution Criteria (measure)
Funding Schedule
Acquisition Management
Program Mgt - Contractor Relations
Project Social / Dev Env
Prog Mgt Structure
Manning at program office
Scope Responsibility
Standards/Certifications
Supply Chain Vulnerabilities
Information sharing
PO Process Performance
Sustainment Issues
Contract Award
Production Quantity
Data Ownership
Industry Company Assessment
Cost Estimate
Test & Evaluation
Contractor Performance
Size
Project Challenge
Product Challenge
Totals
0
Below diagonal
0
Causes
Mission / CONOPS
Effects
0
0
0
0
0
0
1
3
5
0
0
1
0
2
2
2
1
2
5
2
4
3
0
0
0
0
0
0
0
0
0
0
0
0
0
25
26
27
28
29
2.Reduce complexity
of Cause and Effect
relationships via
matrix techniques
3. Assign
Conditional
Probabilities to
BBN Model
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Capability Definition is
affected by CONOPS and
Strategic Vision
30
2.Reduce complexity
of Cause and Effect
relationships via
matrix techniques
3. Assign
Conditional
Probabilities to
BBN Model
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Equation
method
Derived from
calibration
exercise
Engineering Change Proposal schedule delay = .56 + .0323 (waveform) + .0323 (Program Planning) + 0323 (Acquisition
Contracting) +.0215 (technology in motion) + .0323 (Contractor Program Mgt)
31
32
3. Assign
Conditional
Probabilities to
BBN Model
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
An example scenario
with 4 drivers in
nominal state
33
Example: BBN
Connection to the
COCOMO
estimation model.
Drivers
XL
VL
6.20
5.07
7.07
5.48
7.80
4.96
4.05
5.65
4.38
6.24
0.49
0.60
2.12
1.59
1.43
1.62
1.33
1.30
1.43
0.83
0.95
0.87
1.26
1.12
1.10
1.14
PREC
FLEX
RESL
TEAM
PMAT
RCPX
RUSE
PDIF
PERS
PREX
FCIL
SCED
N
H
VH
Scale Factors
3.72 2.48 1.24
3.04 2.03 1.01
4.24 2.83 1.41
3.29 2.19 1.10
4.68 3.12 1.56
Effort Multipliers
1.00 1.33 1.91
1.00 1.07 1.15
1.00 1.29 1.81
1.00 0.83 0.63
1.00 0.87 0.74
1.00 0.87 0.73
1.00 1.00 1.00
XH
0.00
0.00
0.00
0.00
0.00
2.72
1.24
2.61
0.50
0.62
0.62
Product Project
<X>
<X>
<X>
<X>
<X>
X
X
X
<X>
<X>
<X>
<X>
34
2.Reduce complexity
of Cause and Effect
relationships via
matrix techniques
Understand
and analyze cost
model input factors
COCOMO Parameter
Scale Factors
PREC
FLEX
RESL
TEAM
PMAT
Effort Multipliers
PERS
RCPX
PDIF
PREX
FCIL
RUSE
SCED
3. Assign
Conditional
Probabilities to
BBN Model
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
35
2.Reduce complexity
of Cause and Effect
relationships via
matrix techniques
3. Assign
Conditional
Probabilities to
BBN Model
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
BBN Outputs
4
Mapped
COCOMO
value
QUELCE October 2014
2014 Carnegie Mellon University
36
Summary
QUELCE includes the effects of uncertainty in the resulting estimate by:
Making visible the quantified uncertainties that exist in basic assumptions.
Calculating uncertainty of the input factors to the model rather than adjusting
the output factors.
Using scenario planning to calculate how specific changes might affect
outcomes.
37
Selected References
Air Force Cost Risk and Uncertainty Metrics Manual (CRUAMM) 2013
https://www.ncca.navy.mil/tools/csruh/CRUAMM_Printable_Version_16Nov2011_Edited_for_T
he_Joint_CSRUH_05Apr2013.pdf
Design structure matrix methods and applications, Steven D. Eppinger and Tyson R.
Browning, The MIT Press May 2012.
Enhanced Scenario-Based Method for Cost Risk Analysis: Theory, Application, and
Implementation, Paul R. Garvey, Brian Flynn, Peter Braxton, Richard Lee, Journal of Cost
Analysis and Parametrics, 5:98142, 2012
GAO Cost Estimating and Assessment Guide: http://www.gao.gov/assets/80/77175.pdf
Investigation into Risk and Uncertainty: Identifying Coefficient of Variation Benchmarks for Air
Force ACAT I Programs, Shaun T. Carney, Captain, USAF, AFIT-ENV-13-M-05, Air Force
Institute Of Technology, Wright-Patterson Air Force Base, Ohio, Mar 2013.
http://www.dtic.mil/dtic/tr/fulltext/u2/a579314.pdf
Joint Cost Schedule Risk and Uncertainty Handbook
https://www.ncca.navy.mil/tools/csruh/CSRUH_Printable_Version_16Jul2013.pdf
38
Selected References
Principles of Uncertainty, Joseph B. Kadane Chapman & Hall/CRC Texts in Statistical Science,
2011
Risk Assessment and Decision Analysis with Bayesian Networks, Norman Fenton, Martin Neil,
CRC Press, Nov 7, 2012
Software Development Cost Estimating Handbook Volume I, Naval Center for Cost Analysis
and Air Force Cost Analysis Agency, Sept 2008 https://acc.dau.mil/adl/enUS/323892/file/46968/SW%20Cost%20Est%20Manual%20Vol%20I%20rev%2010.pdf
Statistical Methods for Eliciting Probability Distributions, Paul H Garthwaite, Joseph B
Kadane & Anthony O'Hagan, J American Statistical Association
Volume 100, Issue 470, 2005 pp. 680-701.
The Correct Use of Subject Matter Experts in Cost Risk Analysis, Coleman, Richard L. ;
Braxton, Peter J. ; Druker, Eric R.,, 7th Annual Acquisition Research Symposium, May 2010
http://www.dtic.mil/cgi-bin/GetTRDoc?Location=U2&doc=GetTRDoc.pdf&AD=ADA530603
The Design Structure Matrix (DSM), www.dsmweb.org
Uncertain Judgments: Eliciting Expert Probabilities, Anthony OHagan, et.al., Wiley, 2006.
QUELCE October 2014
2014 Carnegie Mellon University
39
Extra
40
Two challenges
Collect the history and classify it
Assist experts to search and identify potential change
DoD
Cloud
Repositories
DoD
Repositories
ARJ
Articles
DoD
Experts
ARJ
Articles
DoD
Experts
CAPE and
Service Cost
Centers
CAPE
and
Service Cost
Centers
Experiment to Test
Experiment
to Test &
Inter-Rater
Reliability
Inter-Rater
Reliability &
Value of Repository
Value1 of Repository
Group
Pretest
Use
DRR
Posttest
2
3
4
41
1. Identify
Change
Drivers &
States
3. Assign
Conditional
Probabilities to
BBN Model
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Information Cloud
Program Rpts:
SARS, DAES
Program Artifacts:
AoAs, ISPs, CBAs
MDAP Data
Sources
DoD
Repositories
DoD
Experts
ARJ
Articles
CAPE and
Service Cost
Centers
42
1. Identify
Change
Drivers &
States
Program Change
Repository
Prog
DDG51
JTRS
F22
State
cond 1
Driver
CONOPS
cond 2
cond 3
cond 1
cond 2
cond 1
cond 2
cond 3
System De
CapDef
InterOpera
Prod uctio
Contract
Functional
CONOPS
3. Assign
Conditional
Probabilities to
BBN Model
Alternative States
Additional
Production
Scope Reduction
deliverable (e.g.
downsized
(funding reduction)
training & manuals)
Program
New echelon
becomes Joint
Trade-offs
Subtraction
Variance
[performance vs
affordaility, etc.]
Obligated vs.
FFRDC ceiling Funding change for Funding spread
allocated funds
issue
end of year
out
shifted
Advocate
Service owner
Senator did not Change in senior
requires change
different than
get re-elected
pentagon staff
in mission
CONOPS users
scope
Scope
Definition
Mission /
CONOPS
Stable
Users added
Additional
(foreign)
customer
defined
New condition
New mission
Capability
Definition
Stable
Addition
Funding
Schedule
Established
Stable
Advocacy
Change
Closing
Technical
Gaps (CBA)
Selected Trade
studies are
sufficient
Technology is
too expensive
Selected solution
cannot achieve
desired outcome
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
Technology not
New technology not
performing as
testing well
expected
~~~~
~~~~
5. Monte Carlo
Simulation to
Compute Cost
Distribution
For C2 systems,
how often does
Strategic Vision
change?
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
43
1. Identify
Change
Drivers &
States
Program Change
Repository
Alternative States
Additional
Production
Scope Reduction
deliverable (e.g.
downsized
(funding reduction)
training & manuals)
Program
New echelon
becomes Joint
Trade-offs
Subtraction
Variance
[performance vs
affordaility, etc.]
Obligated vs.
FFRDC ceiling Funding change for Funding spread
allocated funds
issue
end of year
out
shifted
Advocate
Service owner
Senator did not Change in senior
requires change
different than
get re-elected
pentagon staff
in mission
CONOPS users
scope
Scope
Definition
Mission /
CONOPS
Stable
Users added
Additional
(foreign)
customer
defined
New condition
New mission
Capability
Definition
Stable
Addition
Funding
Schedule
Established
Stable
Advocacy
Change
Closing
Technical
Gaps (CBA)
Selected Trade
studies are
sufficient
Technology is
too expensive
Selected solution
cannot achieve
desired outcome
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
Technology not
New technology not
performing as
testing well
expected
~~~~
~~~~
Effects
Causes
Mission / CONOPS
Change in Strategic Vision
Capability Definition
Advocacy Change
Closing Technical Gaps (CBA)
Building Technical Capability & Capacity (CBA)
Interoperability
Systems Design
Interdependency
3
3
2
1
1
1
Interoperability
System De
CapDef
InterOpera
Prod uctio
Contract
Functional
CONOPS
Advocacy Change
cond 2
cond 3
cond 1
cond 2
cond 1
cond 2
cond 3
Capability Definition
F22
Driver
CONOPS
If Strategic Vision
changes, what
else changes?
JTRS
State
cond 1
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Mission / CONOPS
Prog
DDG51
3. Assign
Conditional
Probabilities to
BBN Model
2
2
44
1. Identify
Change
Drivers &
States
3. Assign
Conditional
Probabilities to
BBN Model
When
both Strategic Vision & Mission/CONOPs
experience change, the BBN calculates that
Capability Definition will also change
Stable
Addition
Funding
Schedule
Established
Stable
Advocacy
Change
Closing
Technical
Gaps (CBA)
Selected Trade
studies are
sufficient
Technology is
too expensive
Selected solution
cannot achieve
desired outcome
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
~~~~
Technology not
New technology not
performing as
testing well
expected
~~~~
~~~~
Causes
Mission / CONOPS
Change in Strategic Vision
Capability Definition
Advocacy Change
Closing Technical Gaps (CBA)
Building Technical Capability & Capacity (CBA)
Interoperability
Systems Design
Interdependency
3
3
BBN Model
2
1
1
1
Interoperability
Capability
Definition
Effects
New mission
Additional
(foreign)
customer
New condition
Advocacy Change
Users added
defined
Capability Definition
Stable
Alternative States
Additional
Production
Scope Reduction
deliverable (e.g.
downsized
(funding reduction)
training & manuals)
Program
New echelon
becomes Joint
Trade-offs
Subtraction
Variance
[performance vs
affordaility, etc.]
Obligated vs.
FFRDC ceiling Funding change for Funding spread
allocated funds
issue
end of year
out
shifted
Advocate
Service owner
Senator did not Change in senior
requires change
different than
get re-elected
pentagon staff
in mission
CONOPS users
scope
Scope
Definition
Mission /
CONOPS
Mission / CONOPS
5. Monte Carlo
Simulation to
Compute Cost
Distribution
Joint conditional
probabilities
Driver State Matrix
can be calculated for
downstream changes.
4. Calculate Cost
Factor
Distributions for
Program Execution
Scenarios
2
2
45
http://blog.sei.cmu.edu
Improving the Accuracy of Early Cost Estimates for Software-Reliant Systems, First in
a Two-Part Series
A New Approach for Developing Cost Estimates in Software Reliant Systems, Second
in a Two-Part Series
47
Contact Information
Presenters / Points of Contact
SEMA Cost Estimation Research
Group
Robert Ferguson
rwf@sei.cmu.edu
Dennis Goldenson
dg@sei.cmu.edu
Jim McCurley
jmccurle@sei.cmu.edu
Robert Stoddard
rws@sei.cmu.edu
Dave Zubrow
dz@sei.cmu.edu
U.S. Mail
Software Engineering Institute
Customer Relations
4500 Fifth Avenue
Pittsburgh, PA 15213-2612, USA
Web
www.sei.cmu.edu
www.sei.cmu.edu/contact.cfm
Customer Relations
Email: info@sei.cmu.edu
Telephone:
+1 412-268-5800
SEI Phone:
+1 412-268-5800
SEI Fax:
+1 412-268-6257
QUELCE October 2014
2014 Carnegie Mellon University
48