Free Resource
A Creative Testing Framework for Paid Social
How to structure a creative test so the account can attribute a result to a decision, and plan the next batch from it.
Planning guide Written by the studio from its own practice; where it states a view, the view is the studio's. It cites no third-party research and attributes no performance figures to the studio.
Most creative tests fail before they launch, because the variations differ in several ways at once and nothing can be learned from the winner. This framework is the structure Unreal writes into a production scope when a brand asks for performance creative: a grid of isolated variables, a fixed control, a naming convention the account can read, and a decision rule agreed before the first impression. It makes no claim about what will win. It is a way of finding out. How large a batch should be is a separate question, answered in how much ad creative you actually need.
What's included
- A question per batch
- One variable per variation
- A deliverables grid as scope and launch plan
- A naming convention the account can read
- A decision rule agreed before launch
- A loop from reads into the next batch
Start with the question, not the concept
A test exists to answer a question the account cannot answer from what has already run: does a problem-led opening hold attention better than a product-led one; does a creator-style piece do what the cinematic cut does; does a price-led offer change the response. Write the question down first. A batch that is not built around a question is a set of ads, not a test.
Isolate one variable per variation
The variables in ad creative are few and well known: the hook (the first seconds), the opening frame, the angle (the argument the piece makes), the offer, the format (creator-style, product film, cinematic, static), the length and the ratio. A variation should change exactly one of them against the control. When two change together, the result belongs to neither.
Plan the batch as a grid
Lay the batch out as rows and columns: concepts down, variables across, one cell per deliverable, with the ratios each placement needs. The grid is the production scope and the launch plan at once. It also shows the batch's size before anything is made, which is where scope and budget get agreed rather than discovered.
Name files so the account can read them
A naming convention that carries concept, variable, variant, format and ratio (for example concept-02_hook-B_ugc_9x16) lets whoever runs the media see the structure in the ad name and report on it without a lookup table. Agree the convention before delivery so the files arrive ready to launch as a structured test.
Agree the decision rule before launch
Decide in advance what counts as a read: the metric the question is about, the minimum spend or delivery the account needs before a difference means anything, and who makes the call. Reading a test early, on the metric that happens to look good, is how a batch teaches the wrong lesson. No universal threshold is offered here; it depends on the account, the objective and the platform.
Feed the next batch
The value of a structured test compounds. Each batch answers its question and raises the next one: the hook that held becomes the control, the angle that lost is retired, the format the account never tried gets its own row. A production partner that can see the reads (read-only, revocable reporting access is enough) can build the next batch from them rather than from a guess.
What the framework does not promise
It does not predict a winner, and it does not replace the media plan. Performance depends on the campaign, the audience, the offer and the account's history, and a good structure makes results legible rather than good. That is also why Unreal never guarantees advertising performance: the scope is written around deliverables and the structure of the test, not around an outcome.
In practice
The service this guide supports: performance creative produced as test batches.
Frequently asked questions
What is creative testing?
Running deliberately varied versions of an ad against a control so the account can attribute a difference in outcome to a specific creative decision, then using that read to plan the next batch.
How many variations should a creative test have?
As many as the account can read against its control. That depends on budget, audience size and the platform's delivery, not on a fixed number, which is why the batch size is agreed in the scope rather than assumed.
Which variable should be tested first?
The one the account has the least evidence about and the most to gain from. For many accounts that is the hook, because the opening seconds decide whether the rest of the piece is seen at all, but the framework starts with the question, not with a default.
Does Unreal run the tests?
No. Unreal produces the batch, structured and named so it can be launched as a test, and hands it to whoever runs the media. It never logs in to an ad account. Where reading results is part of an engagement, reporting access is read-only and revocable.
Start with one scoped batch.
Book a free 15-minute discovery call. We'll map the exact plan for your brand — no pitch deck, no fluff.