Verification Cost and Realised Productivity Gains

Abstract

We model realised productivity gain as a function of generation cost, verification cost and error rate, and fit it against observed outcomes in software organisations. Gains are largest where verification is cheap and shrink toward zero where correctness requirements are strict. The model predicts which domains will absorb capability gains rather than realise them.

Key findings

• 01Realised gain is bounded by verification cost, not generation capability
• 02Gains approach zero in strict-correctness domains
• 03The model predicts observed cross-industry variation
• 04Improving verification dominates improving generation for realised value

Limitations

Published at equal prominence to the findings. A paper’s limitations are usually the part that determines how much its result should move your beliefs.

• Fitted on software organisations only
• Verification cost is difficult to measure directly
• Model assumes error rates are independent of task difficulty