Case Study

From Guesswork Ad Creative to a Weekly Ad System That Learns What Actually Sells

A system that finds where the sales are being lost, creates new ads every week, and learns from what actually sold, not what got clicked.

← All case studies

The Situation

A Facebook ads agency came to me because the ads they ran for their clients were not turning into enough sales, and they could not always tell why:

  • Ads were getting clicks, but the sales were not following, and nobody knew where they were being lost
  • New ads were made on gut feel, with no clear link to what had actually sold before
  • Winning ads ran until they wore out, with no quick way to make fresh versions of them
  • AI ad tools wrote generic copy and made claims their clients could not back up
  • Hours every week went into pulling reports and deciding which ads to pause, for every client

They did not need more ads. They needed to know what was working for each client, and more of that.

What I Built

A system that runs every week for each client, from reading the results to delivering the next batch of ads ready for review.

The Diagnosis

It reads a client's ad account and works out where sales are really being lost: the ads themselves, the page people land on after they click, or the tracking. So the agency fixes the real problem instead of just making more ads.

The Brief Writer

Turns the diagnosis into creative briefs written in the client's own brand voice, each one aimed at where the buyer is in their decision, from first hearing about the offer to being ready to buy.

The Ad Creator

Writes the copy and creates the image for every brief, then puts each finished ad in one place for the agency to review. Before anything reaches them, every ad is checked for claims the client cannot back up, and anything that fails is fixed or flagged.

The Winner Remix

Takes the ads that actually sold, keeps what made them work, and changes one thing at a time, like the hook or the headline, to create fresh versions before the winners wear out.

The Performance Judge

Matches every ad to its results and judges it on recent sales, not clicks or other numbers that look good but do not pay. What it learns feeds straight into next week's ads, so every batch builds on the last.

The Infrastructure

The agency sets the strategy and reviews each batch, and the system handles the week-to-week decisions in between. Every run has a spending limit it cannot go over, and anything serious, like a budget problem or a whole batch failing its checks, stops the run and goes to a person. One bad ad never stops the rest of the batch; it is simply flagged for review. And every step hands its work to the next in a checked format, so nothing gets lost or garbled along the way.

The Results

Weekly
A fresh batch of ads, built on what actually sold.
Every ad
Checked for claims the client cannot back up.
Sales
Not clicks: every ad is judged on the results that pay.
Every run
Has a spending limit it cannot go over.

What Comes Next

Every week the system knows a little more about what sells for each client, so the ads get sharper the longer it runs, and taking on a new client does not mean starting from scratch. The same approach works for any business spending on ads: stop guessing, find where the sales are really lost, and make more of what already works.