Demo mode · in-memory fixtures · swap to Supabase before production
ADMIN · PPS
Configure: Physician Practice Specialists
healthcare · America/New_York
Experiment Mode
Run a 50/50 split between engine-generated content and a hand-written baseline, then read the lift with a 95% Wilson confidence interval. This is the renewal argument: “+X% over baseline, 95% CI.” Outcomes flow in once the publish loop (S6-6) is live; Start is inert until then.
Meta carousel engine vs baselinedraft
metric: click
Engine 50%Baseline 50%
Not started — preview the 50/50 split above, then start.
LinkedIn engine vs baseline — Q3running
Engine-generated LinkedIn posts lift the reply rate over hand-written baseline posts.
metric: reply
Engine 50%Baseline 50%
Engine-generated
46.7% (28/60)
Baseline (hand-written)
20.0% (12/60)
95% Wilson interval · 120 observed of 120 assigned
+26.7pp (+133% relative)Engine wins (95%)