rehearsal

Rehearsal stress-tests a small brand's monthly marketing plan before the money is spent. The problem: a small brand owner with no marketer plans a month of marketing, spends a few hundred dollars, and learns in week 3 that it did not work. The plan was never tested against anything going wrong. How it works: the AI proposes, a deterministic engine tests, the owner decides. Muse Spark reads the owner's profile, proposes four directions written for that brand with parameter ranges, and plays the adversary that invents 20 bad months (creators who never post, a rival's sale, ad costs up 60%, the holiday lift that never comes). It never writes an order count, a revenue figure or a rate. A deterministic Go engine turns the directions into 240 plans, runs each against every bad month, and reports what survived, what each plan dies if, the one assumption that flips the ranking, and where every dollar goes. Two picks also get a combined plan with the budget split. The owner commits, and Launch creates a paused campaign in a real Meta ad account plus a creator seeding kit and a calendar. Before any plan is shown, the engine backtests itself on the brand's own history and shows its hits and misses; it reviews the month that just ended and re-learns; levers with no history are marked "not rated" instead of guessed; Spark's channel picks are checked against public data. Three synthetic demo brands (jewelry, apparel, skincare) run end to end. Live: https://rehearsal-run.fly.dev