← Systems05 of 06

Market Proof

Before anyone writes a script: proof of what's already working in the market.

Less in. More out.

When a new client signs, it searches the ad library for their competitors and for businesses next door to theirs, and finds the ads that have kept running. Here's what that adds up to.

370+winning ads analysed, tagged and transcribed.
10past clients backfilled into the library in one day.
0.2¢roughly what each ad search costs.
0scripts written by AI. People write those.

How it works

When a new client signs, it searches the ad library for their competitors and for businesses next door to theirs, and finds the ads that have kept running. Long-running ads are the ones making money.

Each one gets downloaded, transcribed and broken down frame by frame. The patterns become a brief for the writer: which formats work, how they're structured, and reference ads to watch. A human writes the scripts.

before → now
Before
  • Formats picked on instinct
  • Briefs with no market evidence
  • Copying whatever went viral
  • Research redone for every client
Now
  • Formats backed by what's still running
  • Reference ads inside the brief
  • Structure, not copy-paste
  • A library that grows with every client

The map

Nine steps from a signed client to a writer with proof in hand.

One new client, start to finish

People Sorter Rulebook AI team
01

Who's on it

The library compounds

Every client's research goes into one shared library, tagged by industry, hook, angle and format.

Once a month it also checks which of our own ads are still running, so the winners get tagged and the next client starts from a bigger, better library.

every client makes the next one faster
01Research a client

New ads analysed and tagged.

02Into the library

Searchable by industry, hook and format.

04Next client starts ahead

Proven references on day one.

03Check our own ads

Monthly: which of ours are still running.

better
every
lap

Rules it follows

  • Days running is a hint, not proof.
  • Count unique videos, not copies of the same ad.
  • Never claim a number the data doesn't support.
  • Research only. People write the scripts.
  • Suggest hooks only where the research shows a formula fits.