Predicting Software Defects 
Michael Allegra
QA Manager
 GXS

Introduction:

If you ever found yourself being overwhelmed by the idea of measurement, this is the session for you. Very often we face the question: what evidence do we have that measures are collected and analyzed in a consistent, and repeatable way? Too often, when people are measuring, they don't use the same processes and procedures. In other words, every time when we don't specify how the measures are collected, and analyzed, we leave room for assumptions and interpretations, and such, for errors in our analysis. The fundamental basis for process improvement is to have accurate, and reliable measures. The measurement construct is a simple but very powerful concept that addresses this problem.

Participants will learn:

Learn how to predict the number of defects in a release, the expected rate of defect discovery and estimated defects remaining when you stop testing. This information provides another valuable input to your decision making process on when your software is ready for your customers.

Outline:

  • Introduction to Software Reliability Modeling
    • Origins
    • Basic concepts without the math
    • Benefits
  • Prediction Models
    • Static
      • Traditional method based on historical metrics
    • Dynamic
      • Focus of this presentation, adjusts during software development
  • Prerequisites
    • Describe project types most suitable for this technique
    • SDLC methodologies
    • Tool requirements, basic skills needed.
  • Tool setup
    • Free tool download from JPL
    • Tool configuration steps
  • Walkthrough process
    • Create simple data file to record daily defects
    • Import into tool
    • Setup and run prediction model
    • Extract results into spreadsheet with supplied template
  • Interpret model results
    • What do the numbers mean?
    • Key decision points for adjusting test coverage
    • Use Predicted Failure rate to increase or decrease test schedule
    • Use Total Estimated Defects to show how many defects may be found by your customers
  • Actual experiences
  • Pros/Cons
  • Conclusion

Biography:

Mr. Allegra's 16 year career has spanned many roles related to QA in IBM Global Services division and more recently at global B2B leader, GXS. The last 8 years have involved QA leadership and management positions in e-commerce, utility computing and global supply chain management applications. He has received several awards at IBM for modernizing processes, tools and metrics across engineering and testing organizations. He studied Defect Prediction and Reliability Growth Modeling from experts in IBM Research and applied it to several e-commerce and B2B QA projects. The uses of these techniques were critical in go/no-go decisions for major software releases of global applications.