Concept testing vs iteration: which should we be doing more of?
By Dan BartleyUpdated 3 minute read
The short answer
Test new concepts until you have a winner, then iterate on it, and keep new concepts coming as winners fatigue. A concept is a new reason to buy, for a new kind of buyer, with its own visual; an iteration keeps the idea and changes the hook, format or edit. An account short on winners is often iterating too early.
What is the difference between a concept and an iteration?
A concept is a new idea: a different reason to buy, aimed at a different buyer, with its own visual. An iteration keeps a proven idea and changes how it is told.
| Question | Concept | Iteration |
|---|---|---|
| What changes | The angle, the buyer and the visual world | The hook, the format, the edit or the copy |
| Example | A family-proof sofa for homes with kids and pets, shown in a busy living room | The same idea, opening on a spilled drink instead of the wide room shot |
| What it finds | New buyers and new reasons to buy | A longer life and a lower CPA for a proven idea |
| The risk | Most concepts lose | Changes so small that Meta treats them as the same ad |
When should we test new concepts?
Whenever you lack a winner that can carry more spend. Concepts are how an account finds new buyers, so they come first whenever growth has stalled.
- You have no current winner, or only one
- Your winners are fatiguing and CPA is climbing
- You are launching a new product or entering a new market
- Scaling has stalled because every ad reaches the same buyers
When does iteration pay off?
Once a concept has proven itself at real spend. Iterations then stretch its life and can lower its CPA, because you are improving an idea the market has already accepted.
Iterate the parts that decide whether people stop: the first frame or first three seconds, the opening line and the format. Under Andromeda, change something visible. A new headline on the same frame is likely to be treated as the same ad and to share its audience.
How should we weight each batch?
By the state of the account, not a fixed ratio. The right mix this month depends on whether you have winners worth extending.
| Account state | Weight the batch towards |
|---|---|
| No proven winner yet | New concepts, almost entirely |
| One or two winners carrying spend | Iterations of the winners, plus new concepts every week |
| Winners fatiguing | New concepts first, then iterations of whatever wins |
| Scaling into new buyers | New concepts aimed at buyers you are not yet reaching |
How do we read results for each?
Judge concepts against your control on CPA or ROAS after a fair spend. Judge iterations against their parent, and keep one only if it beats it.
Early signals explain a result: hook rate and hold rate for video, click-through rate for both. They show whether a losing concept failed to stop people or failed to convince them, which decides whether the next version needs a new opening or a new idea. Which creative metrics matter defines each one.
Who comes up with that many concepts?
Whoever owns the testing plan, working from reviews, customer questions and what has already won. The making is usually the constraint, because each concept needs its own visual.
In an e-commerce brand Dan ran creative for, new ads went from under 5 a month to 50 to 100 once each concept no longer needed its own shoot, and the ads got better as well as more numerous.
Other questions people ask about this
Is a new format of the same idea a concept or an iteration?
An iteration of the concept, though Meta sees it as a new creative because the visual changes. Turning a winning static into a video is often the fastest way to extend a winner.
How many iterations of a winner should we make?
A few at a time, each changing one thing, so you know what made the difference. Stop when iterations stop beating the parent.
Does Advantage+ creative do the iterating for us?
Partly. Its enhancements adjust one ad for different people and placements, which helps, but they work from the creative you supply rather than finding new angles.
Still have a question? Ask us