Scaling a winner

Make similar

When something is working, the smart move is to make more like it without starting from a blank brief. The Make similar action does exactly that: it rebuilds a brief from a winning creative, carries that creative's live metrics in as context, and runs the normal generate path to produce new variants that keep what worked. There is also a cron that auto-promotes strong performers and tags them as winners.

Why this is manual right now

Both of those lean on live performance data, and in production there is none yet, because no paid deployments have run. Without real numbers, "winner" has no automatic definition. A restaurant can only put three spins on its best seller next week if the register actually tallied which dish sold out. Right now Pinnora has no register reading. So scaling today is something you drive by hand: you decide a creative is a winner from organic response or your own judgment, then you regenerate around it.

How to do it by hand

The manual version is simple. Take your best ad and generate a new batch that holds its angle and product fixed while varying the hook and the tone. That is the same one-axis habit from earlier in this chapter, pointed at a proven starting point. The ad generator is the surface for it. You are doing by hand what the performance loop will do on its own once a connected ad account is feeding it real numbers.

Be honest with yourself about the signal you are scaling on. Organic likes and gut feel are weaker than paid CTR and ROAS. They are fine for deciding what to make more of, but they are not a closed performance loop, and treating them like one will mislead you.