Diff your best creative against your worst

Reverse the ads that worked and the ads that did not, then find out what actually separates them.

You already paid for this data and have never been able to read it. The difference between your winners and your flops is sitting in the creatives themselves, and nobody has ever written it down.

Why this will not look generic: The comparison is between your own results, so the finding is about your audience rather than about design in general, and it is the one insight no competitor and no model can hand you.

What you walk away with

• A prompt for each winner and each flop, in two folders
• The specific choices that separate them, named
• A checklist for the next creative, built from your own performance

Typical run: 10 winners and 10 flops. Fewer than about six a side and you are reading noise rather than a pattern.

This one happens in one conversation

You bring: Your best-performing creatives and your worst, both reversed and foldered.

Open with: Compare the prompts in my 'winners' folder against the ones in 'flops'. What is actually different, and what is the same in both so I can rule it out?

What the thread holds on to:

• Both sets at once, which is what makes a difference visible rather than a hunch
• What it already ruled out as common to both, so you stop re-litigating it
• The running theory, which you can keep testing as you add more creatives

Then you can just ask:

• What is present in every flop and no winner?
• Is this difference about the image or about the copy placement?
• Write the next brief so it has all four winner traits

What the reverse actually gives you

Point it at: Ten creatives that beat your benchmark, and ten that fell short of it.

What comes back (representative example): The two sets agree on more than they differ on. Both use the product as the dominant element, both sit in your brand palette, and both put the headline top-left, so none of those is doing any work. The separation is narrower and stranger. Every winner has a visible human element, a hand, a shoulder, a reflection, even where the product is the subject. Eight of ten winners are shot in directional light with a clear shadow direction, against three of ten flops. And every flop has three or more competing focal points, where no winner has more than two.

Things you would never have thought to type:

• a human element present in all ten winners, even partial
• directional light in 8 of 10 winners against 3 of 10 flops
• three or more focal points in every single flop
• the brand palette ruled out, because both sets share it

What you change: Nothing about the image yet. The useful output is the rule: a hand in frame, one light direction, never more than two focal points.

How it works

1. Reverse both sets (In your code) - Two folders, one call per creative. The flops matter as much as the winners.
2. File them in two folders - Tag them winners and flops. The folder names are what the next step reasons over.
3. Ask what separates them (In Claude) - What both sets share is as useful as what differs, because it rules things out.

Cost and time

This workflow runs at 2 reverses + about 1 minute. Your result carries automatically from each step into the next, so there is no re-uploading between tools. Reverses are billed from your reverse meter and generation steps from your AI Credits, both bundled in every pay-as-you-go credit pack.

Browse all guided workflows or all individual tools. Pay as you go, no subscription, AI Credits never expire.