Meridian SaaS Platform • Q3 2024 Integrity Audit
Histogram of p-values across 126 completed experiments (Q3) • Healthy distribution should be uniform under null
Weekly SRM incidence rate • Spike in Week 4 due to SDK v3.2 bug
Planned vs. achieved minimum detectable effect across experiment types
Experiments stopped early without correction • By team
Chi-square test results for randomization quality • Green = uniform
Recent experiments flagged for statistical integrity issues
| Experiment ID | Team | SRM Status | P-Value | Sample Size | Power | Issues |
|---|---|---|---|---|---|---|
| EXP-4721 | Growth | SRM Detected | 0.042 | 18,420 | 0.73 | Mobile redirect SRM, Underpowered |
| EXP-4718 | Product | Warning | 0.003 | 42,100 | 0.88 | Stopped early (peeking) |
| EXP-4715 | Platform | Pass | 0.156 | 51,200 | 0.92 | — |
| EXP-4709 | Growth | SRM Detected | 0.001 | 29,800 | 0.81 | Assignment skew, Network contamination |
| EXP-4702 | Product | Pass | 0.089 | 38,500 | 0.95 | — |
| EXP-4698 | Growth | Warning | 0.012 | 22,300 | 0.68 | Underpowered, Collision with EXP-4702 |
| EXP-4691 | Platform | Pass | 0.421 | 47,900 | 0.90 | — |
| EXP-4687 | Growth | SRM Detected | 0.028 | 31,200 | 0.79 | SDK v3.2 bug, Mobile redirect |
Percentage of multi-metric experiments applying Bonferroni/BH correction
Weekly FDR estimate from holdout validation • Target: <5%
User overlap between concurrent experiments • Red = high collision risk