Up
Soft Productivity
Defect Density Requirements?
Defect Containment
Software References

 

Software Costing is confounded by many variables...some are similar to Hardware, as shown in Figure 1.

So.........Why is it Difficult?

  1. Software cannot be examined visually for tolerances which are easily measureable as with Hardware.  Therefore, SW is brittle!!!
  2. As complexity increases, software interfaces can increase almost exponentially...this is not true with Hardware.
  3. Failure in hardware can be resolved by finding the problem in a component which causes the system failure
  4. Current cost models are based on large collections of software data (200 to 1000s of projects), with only qualitative or vicarious judgment on past performance
  5. These efforts lack an understanding of the hardware system problems, the team or the process environment by management reporting the results.
  6. Therefore, these cost models lack fidelity and are not useful to the government or the contractor's management for cost assessment or future estimation.

Enerspace has developed a new approach which demands a minimum set of data to determine software costs for any size program.

Our process requires the following 10 metrics on every program at its conclusion as well on current programs to be reported to customer periodically. 
We have data that has excellent correlation to cost.

  1. Source lines of code by component, module, packages libraries, etc.
  2. Code Change history (added, deleted, modified)
  3. Effort (person-hours)
  4. Schedule to Complete (months)
  5. Hardware Design changes (components, system)
  6. Process Efficiency (Defects and their Lifecycle Containment History)
  7. Head Count History over Lifecycle with Key Personnel Turnover (SW Managers, SW Project Engineers)
  8. System Reliability (with Error types and their Failure Consequences)
  9. Documentation, Meeting and Presentation Requirements
  10. Requirements Volatility (added, deleted, modified)

Reality Checks:
Run Sensitivity Analysis using Monte Carlo Methods with Commercial Databases and Tools to Understand Schedule Impacts...see Figure 2.


Figure 1  Software Cost Estimation

 

12 Month Schedule: $1,370,000

15 Month Schedule: $560,000

 

Figure 2  Contrux Estimate Tool showing Nominal, High and Low Cost Plans affected by Schedule