Bayesian Reliability Predictor (Software)

Bayes' Rule (odds form)
Posterior Odds=Prior Odds×∏i=1kLRi\text{Posterior Odds} = \text{Prior Odds} \times \prod_{i=1}^{k} \text{LR}_i
Posterior=Odds1+Odds\text{Posterior} = \frac{\text{Odds}}{1+\text{Odds}}

Each factor (coverage, reviews, complexity, size, defect density, experience, maturity) contributes a likelihood ratio (LR) that adjusts prior belief to obtain a posterior reliability probability.

Inputs

e.g., 0.999 = ≤1 failure per 1000 hours

0.60
80%
70%

Use cyclomatic complexity bands

Thousands of source lines

Expected pre-test defect density

3

Results

Posterior P(High Reliability)
79.8%
Estimated Residual Defects
16.2
Target R
0.999
Prior: 0.60
Coverage: 80%
Reviews: 70%
Complexity: Moderate
KLOC: 100

Sensitivity (Tornado)

Bars show % change in posterior reliability if each factor is improved to a strong level.

What-if Simulator

Notes & Assumptions