Test and publish

Test formats and openings with generated presenters first, then remake the winners with real creators before ad money goes in.

Test two to five angles, each with its own natural hook, then three or four hooks inside the winning angle.

Start from a post at ten times or more its account's median; recreate its structure with your own material and change one thing.

Each variant gets an id, a hypothesis, the one thing it changes, what stays fixed, a cost, an observation window and the result that would justify the next test.

One body, two to six openings, each posted as an Instagram trial reel; read the views after about 24 hours and compare the batch with itself.

Show someone only the first frame for two seconds and ask what the video is about.

Pay each creator a share of the spend behind their own ads, with a monthly floor for a video quota, and give them the product and the news, not a script.

A plain 9:16 screen recording of someone using the product, 30 to 60 seconds, one line of text on top.

After a breakout, post the next piece on the same promise within 48 hours.

Set the platform's own AI label and its commercial-content setting before the post goes out.

Cut the finished long video first, then clip the shorts from that cut.

Write the account's promise in one line; every post keeps it, in two or three repeatable formats, at one fixed time in the audience's evening.

One sentence the viewer would type into search, holding the words of the account's promise.

Hook rate is three-second views divided by impressions; hold rate is the share that keeps watching past the opening.

Plays divided by the median plays of the account's other recent clips.

Compare reach and engagement with the account's own medians, never with another account.