Creative strategy
Concept versus variant, a testing framework, and the creative metrics worth watching.
What caps your account is not your budget — it is the number of winning creatives you have. However good a single creative is, it cannot carry spend past a certain point; the audience has seen it.
Concept versus variant
This is the most important distinction in test design, and the one most often collapsed.
| Concept | Variant | |
|---|---|---|
| What it is | A different idea, angle or format | A different execution of the same idea |
| Example | "Saves you time" replaced by "Saves you embarrassment" | The same ad with a different opening frame |
| What it teaches | Which message lands | The best presentation of that message |
| Effect | Potential step change | Incremental improvement |
A new concept is unlikely to win, but when it does it carries the account. Variant testing is safe but never raises the ceiling. Do not mix them: find a concept first, then exploit the winner with variants.
Testing cadence
Set testing up as a process, not a campaign. A cadence that works:
| Step | Practice |
|---|---|
| Volume | At least 8–12 new concepts per month |
| Batching | Ship concepts in batches, not one at a time |
| Decision time | 3–4 days per concept, or ~3× target CPA in spend |
| Culling | Kill the bottom half, multiply the top quarter with variants |
| Archive | Record the losers too — why it lost is input for the next round |
A realistic expectation: roughly one in six tested concepts becomes a scalable winner. That rate is not low, it is normal. Which is why the answer is producing more, not guessing better.
Do not kill a concept before it has spent about 3× your target CPA. Culling early means deciding on noise rather than data.
Metrics to watch
Looking at a single outcome metric on video hides where the problem is. Read the funnel step by step:
| Metric | Formula | What it tells you |
|---|---|---|
| Hook rate | 3s views ÷ impressions | Is the opening getting attention |
| Hold rate | 15s views ÷ impressions | Is the content keeping it |
| CTR | clicks ÷ impressions | Is the promise motivating action |
| CVR | conversions ÷ clicks | Does the page deliver on the promise |
The diagnostic logic is simple: low hook means change the first seconds; good hook with poor hold means the content does not carry the promise; good CTR with poor CVR means the problem is not the ad, it is the landing page.
Message angles
Genuinely different angles you can build for the same product:
- Problem–solution — show the pain, present the fix
- Before–after — make the transformation visible
- Comparison — explain why the alternative falls short
- Social proof — real user, real result
- Moment of use — show how the product enters daily life
- Objection handling — address the hesitation blocking purchase head-on
- Curiosity — open with an unexpected claim or image
If your creative library does not cover most of these angles, you are not running creative tests yet — you are running variant tests.
Choosing a format
| Format | Where it is strong | Watch out |
|---|---|---|
| UGC video | Acquisition, trust, low production cost | Over-polished and it reads as an ad |
| Static image | Fast testing, clear offer delivery | Limited storytelling capacity |
| Animation | Visualizing an abstract benefit | Does not convey product texture |
| Catalog | Wide product ranges, retargeting | Not suited to concept testing |
In practice statics are the most efficient way to test a message angle because they are fast and cheap; move the winning angle to video once you have found it.