Refract Group2026
Glossary
Creative testing
Creative testing is the structured practice of running versions of an ad against each other, each made to answer one question about what makes customers act, and keeping what each round taught for the next round of ads.
- Customer questionDo they believe the claim?
- Angles and versionsOne thing changes per pair
- Split test runsAudience and budget agreed
- Enough evidence?Record inconclusive
- Keep, revise or stopBrief for the next round
Creative testing in detail
- Also called
- Ad testing, Ad creative testing, Creative experimentation
- Topic
- Growth
- Updated
What it is
Creative testing is how a team finds out which ideas in its advertising work. A round starts with a question taken from what customers ask or doubt, such as whether they believe a claim or understand how setup works. The team writes the angles that answer it, makes versions of each, runs them and reads the result against an action agreed beforehand.
Two kinds of test sit inside it. A concept test sets complete directions against each other and tells you which to pursue. A variable test changes one part, such as the hook, the proof or the format, and tells you which part mattered. Ad platforms run either through their own experiment tools, which split the audience so each person sees one version.
What makes it a practice is the record. The brief, the files and the result stay together, so the next round starts from the last one.
Why it matters
Without a plan, a paid account fills with ads that differ by accident, and a result cannot be traced back to an idea. Spend goes wherever the platform delivered most, and nobody can say why the best ad worked or what to make next. A planned round costs the same money and leaves something the team can use: the claims people believe, the proof they respond to and the openings that have stopped working.
How we do it
Before anything is made, we write the test plan: the audience, the question, what changes and what stays the same, where each version sends people, the action that counts, and what result would make us keep, revise or stop a direction.
Each angle then gets the versions needed to test one clear difference, such as the opening, the proof or the format, and the number is set in the scope by what you want to learn. Because every ad is assembled from one kit of parts, a later version reuses what is already made. Every file is named by its message and format, so a result can be traced to the ad behind it, and a weak or uneven result is recorded as inconclusive.
An example
Say a security product’s ads promise setup in a day, and sales calls show buyers doubt it. The question is whether showing the setup persuades more than evidence from a customer. Version one is a 20-second screen recording of setup. Version two makes the same claim over a quote from an IT lead. The audience, offer and landing page stay the same. The recording brings more trials that activate, and the next round tests three openings for it.
Common mistakes
- Reading the clicks and stopping there. A striking image can win attention from people who never buy, so judge each version on the action the campaign is for and on the quality of what it brought in.
- Changing three things at once. A new promise, a new proof point and a new format in one version can tell you which ad to keep and cannot tell you which change did it.
- Treating ordinary delivery as a test. When several ads run in one ad set, the platform decides how often each is shown, so they seldom get an equal share or the same people.
- Splitting a small budget twenty ways. None of the versions gets enough activity to judge, and the winner rests on a few conversions.
- Losing the reason behind the file. A folder of ads called “final” cannot tell the next person which question each one was made to answer.
Creative testing vs A/B testing
Creative testing is the ongoing practice of deciding which ideas go into ads and learning from each round, and an A/B test is one statistical method it can use to compare two versions fairly.
| Creative testing | A/B testing | |
|---|---|---|
| What it is | A practice that runs in rounds | One controlled experiment |
| Compares | Ideas, angles and formats | Two versions of one thing |
| Used on | Ads | Pages, products, emails and ads |
| Leaves behind | A brief for the next ads | A winner, or no clear difference |
Questions
What people ask about creative testing.
A creative test should run until each version has enough conversions to judge, which depends on the current conversion rate and the size of difference you need to see. There is no fixed number of days. Google Ads suggests at least four to six weeks for a campaign experiment, and whole weeks keep weekday and weekend audiences in proportion.
Start with the claim, because what an ad promises decides who responds at all. Creative testing on the hook, the format or the colours is worth more once you know which claim works. Each round should change one of those, so the result says which part made the difference.
Creative testing is the whole practice of choosing which ideas go into ads, making versions and learning from each round. A/B testing is a statistical method for comparing two versions with a randomly split audience. A creative test often uses an A/B test, and A/B testing is used on pages, emails and products as well as ads.
Where this shows up in our work
Related terms
Working on this now?
Tell us what you’re building and where it’s stuck. We reply within 12 hours.
Most teams start with a call, then a sprint or an audit.
The people you’d work with, from the first call to launch