How do you test ad creative?

Which creative testing tools do Meta, TikTok and Google offer?

Meta, TikTok and Google each offer a creative test tool. According to Meta, you run an A/B test in Meta Ads Manager to understand which ad variant delivers the best results. Meta's A/B test splits audiences at random.

TikTok's version is the split test. Per TikTok, split testing in TikTok Ads Manager compares two different versions of your ads. TikTok splits the audience into two equal groups, and each group sees only one ad group.

Google Ads offers video experiments. According to Google, the basic option, A/B test video assets, makes creative the single variable in a two-armed test of control against treatment. That option is only available for Video Reach and Video View campaigns. Other video campaign types use a Custom experiment.

Running two ad sets side by side doesn't give an even split on Meta. Per Meta, ad sets that run at the same time without the A/B Testing tool aren't split evenly, because the system treats them in combination. That skews delivery and budget distribution between the two ad sets.

How many creative variants should one test compare?

The number of variants in a creative test depends on the platform's tool. According to Meta, A/B Testing lets you test up to five ad variants. TikTok's split test compares two versions. Google's Custom video experiment can test any variable with up to 10 arms, while A/B test video assets is a two-armed test.

Meta says only one variable should change in an ad test, while the rest of the campaign setup and creative remains identical. TikTok allows only one variable for each split test.

On TikTok, the Creative Assets variable covers images, videos, ad formats and the initial hooks of your video ads. TikTok defines those hooks as the opening 2-3 seconds of a video ad.

TikTok's creative best practices page suggests 3-5 different creatives per ad group and 3-5 diversified ad groups per campaign. The same page says that, as a general rule, it's always better to use creatives with big differences, especially when testing.

How long and how much budget does a creative test need?

TikTok sets a time floor for a creative split test. According to TikTok, you should set split test schedule dates to at least 7 days, so the system has enough time to collect a useful sample.

For budget, Meta says an adequate budget ensures an even split and confident strategy determination in an A/B test.

If you run a Demand Gen experiment with conversion based bidding, Google requires a minimum of 50 conversions per arm to surface results. For conversion related metrics in Demand Gen experiments, Google says you need at least 100 data points to start seeing results. Budget each Demand Gen arm to reach 50 conversions when you use conversion based bidding.

When this doesn't apply: Meta's learning-phase figures here come from its guidance for Horizon Store app campaigns. Per that guidance, learning occurs at the ad set level and stabilizes with a minimum of 50 events over a 7-day period. The same guidance says changes to your ad set can trigger a new learning period.

How do you read the result of a creative test?

TikTok and Google name a creative test winner at different confidence levels. According to TikTok, split testing is designed to perform an A/B test with a 90% confidence rate. TikTok selects and verifies a winning ad group only if the results are statistically significant.

Google Ads video experiments use a sliding scale. Google recommends waiting until the experiment finishes for conclusive results. Directional results can show as early as a 70% confidence threshold, 80% confidence is still directional, and 95% at completion is conclusive.

On TikTok and Google, a lower cost in one arm isn't a win until the tool reports a winner at its confidence level.

In Demand Gen experiments, Google reports Similar performance when there's no significant difference between arms at the chosen confidence level. Google says an experiment without a winner might mean the creative variation you tested isn't a substantial one. It suggests testing other asset types or more significant variation in the next experiment.

When is a formal creative test not worth running?

If a Demand Gen arm can't reach 50 conversions under conversion based bidding, the experiment won't surface results. According to Google, Demand Gen experiments with conversion based bidding require a minimum of 50 conversions per arm to surface results.

Meta lists three occasions to run a creative test: when you have a hypothesis for how you could drive performance improvements, a new creative concept or campaign setup, or a question or see performance declines.

A creative test also has a scope. Per Meta, A/B Testing can help you quickly understand which strategies work best and make tactical optimizations. Conversion Lift helps you measure the conversions directly caused by your ad, which Meta says makes it useful for strategic decisions like where to increase your budget.

When this doesn't apply: Google states the 50-conversion minimum for Demand Gen experiments only. On TikTok, the tool selects and verifies a winning ad group only if the results are statistically significant.