Hook Variants Built to Falsify a Positioning Claim

Twelve openers come off a single shoot block, but four is the number that decides whether the test is worth running: four genuinely different hypotheses about why a stranger stops scrolling. Everything past four is the same idea in a different jacket, and a test full of near-duplicates returns a flat read that costs you a month of media. We would rather sell you four ideas shot twelve ways than twelve ideas that turn out to be one. What counts as a hypothesis here: Not warmer colour. Not faster cutting. A hypothesis names the reason someone keeps watching: the objection answered in the first second, a person on screen who is recognisably the viewer rather than the brand, a demonstration that looks unstaged, the price said out loud before anything else, the problem shown before the product exists in frame. Each opener is built so that if it wins you learn something applicable to the next twenty things you make, including the ones we do not make. Shooting for the format instead of cropping into it: Vertical is framed vertical on the day. TikTok and Reels lay interface over the top and bottom of the frame, so composition leaves that room rather than discovering it at export. On-screen text is set to be legible at the size the feed actually plays, on a phone, at arm's length, which we check on a phone before delivery rather than on a colour-managed monitor in a dark room. Sound is mixed for a device speaker, and every variant also has to work silent, because a large share of the impressions you buy will be exactly that. What we report and what we refuse to promise: Hook rate as three-second plays over impressions, hold at six and fifteen seconds, and the drop-off frame per variant. We will not promise a CPA and we will not accept a brief that makes creative accountable for one on its own. What we insist on instead is a test that can be read. A cell needs enough conversions behind it before a difference means anything, and if the budget will not fund a readable result across twelve cells we shoot fewer variants and say so on the first call. Shipping variants an account cannot fund is how a shop bills a month and a client learns nothing. What we refuse: No synthetic faces or cloned voices in anything a viewer is meant to trust. No staged footage dressed up as user-generated. No opener aimed at children that asks a child to ask a parent to buy, because the rules on direct exhortation apply regardless of whether the ad ran on television, and platform enforcement is not the standard we work to. We also do not manage creator rosters. We make the films. Your creator relationships stay yours. Who should not hire us: Teams without a media owner able to restructure the account so the test actually runs cleanly. Teams spending under roughly fifteen thousand a month on the channel, where the honest advice is that targeting and landing page are still the constraint and creative is not. Teams who want fresh variants every week indefinitely, which is a content factory and there are good ones in Warsaw that cost less than we do. How it runs: A working session with your media lead where we go through what has already run and what the drop-off frames are telling you. Then the hypothesis set, agreed and written down so it can be judged later. One shoot block, cast for the format rather than for the brand film. Edit, versioning, delivery to platform specs with captions both burned and supplied as sidecars. Two to three weeks after launch, a read session on the results and one decision: which hypothesis earns a second round and which is dead.