Hook Sets Built as Hypotheses, Read on Hold Rate

A hook test needs roughly 8,000 impressions per variant before three-second hold rate stops moving. That number governs this entire engagement. If your paid social budget cannot push a dozen openers past it inside a week, you do not have a test, you have a preference survey with a spreadsheet attached. We read your account before quoting and tell you how many hypotheses it can actually resolve. It is usually fewer than the number in your brief. Hooks are hypotheses: Every opener names the mechanism it tests and predicts a direction before it runs. Price stated in the first two seconds against price withheld. Objection first — the reason people do not buy, said out loud — against benefit first. Product in frame at zero against a held human face. A line that works sound-off against one that needs audio. Native-format mimicry, which for a furniture brand means the opener looks like somebody's actual flat and for a dealer group means it looks like a walkaround filmed on a phone. Openers differing only in font, colour or track are not variants, because whatever wins there you cannot repeat. What we read, and what we ignore: Three-second hold, and six-second where the placement reports it, taken from Meta and TikTok retention curves per placement rather than blended. Click-through is downstream, influenced by the hook and dominated by the thumbnail and the offer, and we do not call hooks on it. A hook is called when the hold-rate gap survives two consecutive days at a spend where every variant has cleared the impression floor. Losing hooks: They carry more information than the winner. A set where every human-face opener lost is a finding about your account rather than about those openers, and it changes what gets cut next cycle. The read-out includes what the set ruled out, in writing, including the cycles where nothing separated — three of the last twelve. Declined on principle: The gimmick interrupt: the unrelated jump cut, the fake glitch, the manufactured double-take. It buys a hold-rate point and poisons everything downstream of it. Actors performing as satisfied customers. Trending audio used as the hook itself, because the finding expires when the sound does. And we will not cut one more opener to round the number up for a deck. Beyond the brief: The body and ending of the ad, unless you ask separately — a hook set is designed to bolt onto creative you already run. Media buying: we take read access and decline spend authority. Landing pages. Creative for placements you are not already buying, where there is no baseline to read against. Say no to us if: Your paid social spend is under roughly €20,000 a month, where the test cannot resolve and the honest advice is to put the fee into media instead. You have one evergreen asset you would rather not disturb. Your legal review takes three weeks per asset, because cadence is the entire mechanism and a hook set that ships in six weeks documents your approval process rather than your market. Cadence: We start inside your ad account, reading twelve months of creative against its retention curves. Then a hypothesis sheet you sign off, which is short and is the actual argument. Production runs from your existing footage plus inserts, or from a capture half-day we arrange in Istanbul or Bursa when the archive has run out. Delivery is per placement, framed and captioned correctly, using the naming convention already in your account so your reporting stays readable.