COSYSMO Usage Experience Panel: What is Happening at Lockheed Martin

1
COSYSMO Usage Experience Panel What is
Happening at Lockheed Martin

• Garry Roedler, Lockheed Martin Engineering
  Process Improvement Center
• garry.j.roedler_at_lmco.com
• Pete McLoone, Lockheed Martin Integrated Systems
  and Solutions
• peter.j.mcloone_at_lmco.com

2
Talking Points

• Collecting data points for the industry COSYSMO
  Effort across Lockheed Martin Business Areas
• Electronics Systems has completed a usage pilot
• Integrated Systems Solutions has done some
  preliminary evaluation of the model
• Lockheed Martin is deeply involved in extending
  COSYSMO to Cover Risk and Schedule, distributing
  effort across the life cycle, and addressing
  Reuse
• Intending to do more local calibration and
  evaluation
• Promoting COSYSMO internally via Lockheed Martin
  Measurement Workshop and other communications
  mechanisms
• Contributing to the review of the COSYSMO User
  Manual and other aspects of the industry effort

3
Collecting Data

• Reasonable success obtaining data points need to
  capture many more
• Starting to obtain data relevant to the COSYSMO
  Risk Module
• Difficult to validate effort and size data
  without a lot of effort
• Suspect very uneven methods for counting
  interfaces
• Engineered system requirements seem to work
  better than customer system requirements
• Labor hours are probably somewhat uneven in terms
  of rules utilized cannot easily supply a EIA
  632 breakdown
• Cost drivers are sometimes misinterpreted but not
  a big problem helps if a team of systems
  engineers is used and a COSYSMO expert can
  facilitate their determination

4
Electronic Systems Pilot

• Had to adjust inputs because of reuse did not
  calibrate to local data default parameters
  generated high estimates compared to actual
  effort
• May be useful as a sanity check for SE cost
  estimates
• Reuse and local calibration need to be added
• Academic COSYSMO is bare bones spreadsheet
• Does not have features of other parametric models
• Reuse, local calibration capability, Class of
  cost, Schedule, Risk, ranges for parameters, Cost
  particular phase(s), labor distributions, summary
  reports, graphical outputs
• Commercial version SystemStar has some of these
  features
• Need a comprehensive Users Guide (in
  development)
• Better guidance on how to combine multiple
  questionnaire attributes into a single model
  parameter
• Better definitions of attributes, with examples
• Definitions of phases covered
• Next steps more data points, local calibration,
  use reuse and risk profile and evaluate

5
Integrated Systems and Solutions

• Did not do local calibration
• Results with four data points showed reasonable
  results
• But data collection was quick and dirty so not
  much confidence in the results
• Nonetheless, it appears with local calibration
  the model should be usable
• The road ahead involves more careful data
  collection to obtain sufficient data points to do
  local calibration properly

6
Final Thoughts

• Local calibration is only as good as the data
  that goes into it
• Dont underestimate the effort needed to obtain
  good data
• Experiment!
• Is accuracy related to size driver weights?
• Is accuracy related to cost driver weights?
• Does accuracy get better if I throw away a size
  parameter?
• Provide quality data points to Ricardo

7
(No Transcript)