Pre-registered A/B tests on outbound messaging. Units are assigned by a deterministic hash, each test declares its primary metric and minimum sample up front, and no winner is declared until the sample is reached.
| Experiment | Status | Primary metric | Units | Verdict | Recommendation | Dates |
|---|---|---|---|---|---|---|
| Subject line: question vs statement Campaign: AI agent launch → reliability | running | Reply rate | 83 accounts | Insufficient sample | Keep running | Sep 28, 2026 – running |
| Funding-trigger personalization vs generic ICP messaging Campaign: Series B–D funding trigger | stopped | Positive reply rate | 303 accounts | No significant difference | No change | May 2, 2026 – Sep 11, 2026 |
| Provocative subject line vs plain value subject Campaign: ICP: AI platform leaders (generic) | stopped | Reply rate | 542 accounts | Treatment wins | Do not ship negative reply, unsubscribe breached | Apr 22, 2026 – Oct 18, 2026 |
Verdicts come from a two-proportion z-test (α = 0.05) on the primary metric. “Insufficient sample” means the test has not reached its pre-registered minimum; it is not a result.