Experiments — A/B, honestly

Variants, claims, subjects — same framework, one table. Click any row to drill into the A/B detail below.

Name & segmentControl vs. treatmentSampleMetricWinnerStatus

· 25%

B · · 0%

6 / 40

reply rate

Δ +7pp (early · n=6)

pending

running

· 0%

B · · 0%

3 / 40

reply rate

too early

pending

running

· 100%

B · · 0%

4 / 20

reply rate

Δ +54pp · p<0.05

concluded

· 50%

B · · 0%

4 / 30

meeting rate

Δ +35pp (winner trending)

pending

running

· 40%

B · · 10%

40 / 40

meeting rate

Δ +31pp · p=0.03

concluded

· 80%

B · · 58%

30 / 30

open rate

Δ +22pp · p<0.05 (too credential dropped open rate)

concluded

· 0%

B · · 0%

2 / 30

open rate

too early

pending

running

Signal-led vs Thesis-led (Care) · care

scope: hook · split 0.5 · started jun 12 · 6 / 40 sends · p=0.34· stop: p<0.10, day 21

A · control

care-signal-led · 32-day-clock

year-one board deck + ai-originated demand

sends 4replies 1rate 25%

B · treatment

care-thesis-led · ai-demand-inflection

care.com: the ai-demand inflection

sends 2replies 0rate 0%