On the first Monday of every month the creative team fills in a grid. Four columns, three rows, twelve boxes. By Friday every box is an ad. The grid is the least glamorous thing we own, and it is the reason our clients' results do not collapse when one ad gets tired.
Why not twelve versions of the winner
The tempting plan is to take last month’s best ad and make twelve variations: new colour, new first line, new music. It feels efficient and it tests almost nothing. The platforms treat near-identical ads as the same ad, show them to the same people and wear them out together. When we looked back at our own accounts for 2025, a batch built from one idea lost half its performance in a median of 19 days. A batch built from four different ideas took 46.
So the columns of the grid are ideas, and an idea is a reason to buy, not a look. For a sparkling drink the four might be: it tastes loud, it has no sugar, it replaces the afternoon coffee, it is what you bring to a dinner. Each speaks to a different person, or to the same person in a different mood.
The rows are formats
Each idea is then made three ways. One is a film of fifteen to thirty seconds. One is a still image with a headline. One is made by a customer or a creator in their own words. Different people stop for different things, and the same idea often wins in one format and dies in another.
The Fizzle square shown here came out of the “no sugar” column in June 2026. As a film it was ordinary. As a still, with five words and a very loud background, it had the highest click-through rate in the account that summer at 2.3%. We would not have found that if the idea had only been made once.
An idea is a reason to buy. A new background colour is not an idea.
Where the ideas come from
Not from a brainstorm. In the last week of each month the strategist brings three sources to the table: the fifty most recent customer reviews, the search terms people typed before they bought, and the comments under the current ads. We look for sentences customers say that the brand does not. “I keep one in the car” became an entire column for Fizzle. Nobody in the room would have written it.
One column is always reserved for the idea we think is wrong. Somebody argues for it, we make it properly, and about one month in five it wins. That column keeps us honest about how little taste predicts.
What happens after Friday
All twelve launch the following week on small, equal budgets. After seven days we read the columns before we read the ads: which idea, across its three formats, brought orders most cheaply? The winning idea gets next month’s bigger budget and three new executions. The losing idea leaves the grid and a new one takes its column.
The client sees the grid at both ends. On Monday it is twelve one-line descriptions they can approve in ten minutes. A month later it is the same grid with a cost per order in each box. It is the simplest report we make and the one clients forward most.
If you only take one thing from this: count your ideas, not your ads. Twelve ads from one idea is one test. Twelve ads from four ideas is four, and four is usually enough to find the one you did not expect.




