If you’re the growth head or the co-founder managing growth, write the early scripts yourself. The scripts will make or break this channel, and you need to learn what works before you hand the job to someone else.
Start with the hook. Test the opening line, the text on screen, and the way the creator enters the video. If you have 10 creators, split them into groups and give each group a different hook. This gives you a cleaner comparison than sending everyone a completely different idea.
A larger creator roster also lets you run more experiments. If you’re preparing two script variants a day, you’ll write 14 scripts across a week. Some should build on ideas that have already worked; the rest should search for the next winner.
We use roughly 30 to 40 percent of our scripts for iterations on winners and the remaining 60 percent for new experiments. On a smaller seven-script plan, that might be three versions of a winner and four new ideas.
I call a script a winner when it unlocks views in our niche. Once you find one, resist the urge to move on immediately. Change the hook. Try different on-screen text. Give the same idea to another creator. Keep the part that worked and vary enough around it to understand why people stopped scrolling.
Two or three exceptional creators can help you discover a format. The wider roster is how you test whether that format travels. When something wins, send the reference and the next variation to everyone quickly, while the idea is still useful.
You still need fresh experiments every week. A roster that only repeats its last hit will eventually run out of room. The split keeps the channel moving in both directions: exploiting something that already has evidence behind it and looking for the next format that can work.
Founder or growth-lead involvement doesn’t have to last forever. Once you have several winners, clear brief templates, and a repeatable way to review scripts, bring in copywriters. Give them the playbook, the references, and examples of the iterations that worked. They can now take over a process you have already tested.