Creative testing
Creative testing that asks one question at a time.
Most ad accounts change five things at once and learn nothing. We change one, say why, and decide what happens next.
Built for the placement
One idea, shaped for where it runs.
The same product reads differently in the feed, in a full-screen vertical placement and in a carousel. These are concept examples for a fictional brand, not client ads.
Kesso
Still cold at 4 pm.
An insulated bottle for long days.
Concept example
Stops the scroll with one clear frame. The message has to land without sound.
Kesso
Your 4 pm water, still cold.
Concept example
Full height and fast. Hook and product stay clear of the top and bottom, where the placement adds its own overlays.
Kesso
The 2 pm slump.
Concept example
Kesso
Double wall. No sweat.
Concept example
Kesso
One bottle, all day.
Concept example
A short sequence: the need, the detail, the payoff. Each card has to work alone too.
Ratios here are art direction. Exact placement specifications are checked in Ads Manager before any client asset is produced, because Meta updates them.
The test plan
Written down before anything launches.
If we cannot say what a test would teach us, we do not run it yet.
Test planConcept example
- Hypothesis
- A concrete time-of-day hook will hold more of the right attention than a general benefit line.
- Variable
- The opening line only. Visual, offer and audience stay the same.
- Placement
- Vertical Reels and Stories first. In-feed read separately.
- Question
- Do people who stop go on to view the product and act?
- Review point
- After enough delivery to see a pattern, not after one good day.
What we test
Variables worth isolating
The opening
The first line or first second. The cheapest thing to change and often the most telling.
Proof framing
Demonstration, detail, comparison or explanation. Only proof that is true and can be shown.
Setting
Where and how the offer shows up in someone's day.
Format
Single image, vertical video or carousel, read separately rather than mixed into one result.
Offer framing
Only with your approval, and only offers you actually run.
Reading results
Evidence, read with its limits.
A result means something only in context: the campaign objective, how well tracking works, how much was spent and how long it ran. Platform-reported conversions are a model of what happened, not a full count.
So reviews say what we know, what we suspect, and what we would need to be more sure. The review rhythm is agreed with you at the start and depends on spend and how quickly tests gather delivery.
Reading the signalNo figures shown
- DeliveryDid it reach the people we intended?
- AttentionDid it hold past the opening?
- Qualified actionDid viewers take a step that matters to the business?
- ContextTracking quality, spend level, timing and sample size.
Illustrative layout. Real reviews use the client's own account data.
Decisions
Every review ends with a decision.
Examples of the kinds of decisions a review can lead to:
Keep and iterate
The idea earns another round with a sharper variable.
Retire the angle
Not working for this audience or placement. Stop spending on it.
Rebuild for the placement
A good idea in the wrong shape. Recompose before judging it.
Fix measurement first
If tracking is unreliable, more testing only produces more noise.
Tell us what your Meta campaigns are doing now.
A short form, a few questions about your business, spend and goals. If it looks like a fit, the next step is a conversation, not a pitch deck.