| Risk | Impact | Mitigation | |------|--------|-------------| | Local maximum trap | Miss 10%+ breakthrough | Run quarterly radical A/B test | | False positives due to high velocity | Wasted dev time | Use FDR (False Discovery Rate) correction | | Segment degradation | Gains for some users, losses for others | Monitor mobile vs desktop, new vs returning | | Statistical significance illusion | Implement neutral changes | Require Bayesian posterior probability >95% |
4.5/5