4 Ways to Approach Creative Testing with Meta Advertising

Creative testing with Meta advertising is an inexact science. It seems every advertiser has their own approach. Some will swear it’s the “right way,” but the best option for you is more nuanced.

It depends on the situation, what you care about, and what you’re trying to accomplish.

In this post, we’ll cover four ways to approach creative testing:

  1. Create Multiple Ads in an Ad Set
  2. Use Text Variations
  3. Use Dynamic Creative
  4. Run an A/B Test

By the end, we’ll discuss what I do and how to determine what’s best for you.

1. Create Multiple Ads in an Ad Set

There’s nothing wrong with kicking it old school and simply creating multiple ads for a given ad set with different combinations of copy and creative. But, there are some basics to consider when doing this.

Ads won’t be shown equally, if at all.

If you’re a control freak, this will drive you crazy. Just because you created four ads doesn’t mean that Meta’s ad delivery algorithm will show them equally to help you understand what works best. In fact, one or more of the ads may not show at all.

There has to be a certain amount of letting go of control with this approach. You need to be okay with the fact that you might create an ad that doesn’t get shown. Or maybe your favorite ad won’t get shown the most.

Trust the algorithm, but understand its imperfections.

When you take this approach, you embrace the chaos and imperfection of it from the outset. Your ads won’t be shown equally and some may not be shown at all. You are trusting that the algorithm will use historical and real-time data to help deliver the right versions to the right people.

But the algorithm will also make these decisions very quickly because, in most cases, any differences in ad performance won’t be statistically significant.

This isn’t a true split test.

If you have multiple ads in an ad set, it’s not a true A/B split test. The same user can see more than one version of your ads. In many cases, this is preferred anyway. But, that overlap means that you’re not going to see results based on a true, scientific split test.

And you need to be okay with that.

Consider a limit of six ads.

Assuming you aren’t running an Advantage+ Shopping Campaign, Meta recommends using no more than six ads in an ad set. Once you’ve added more than six, there’s marginal benefit.

Create the ads at initial publication, if possible.

Every time you publish a new ad, you’ll restart the learning phase. Not every advertiser sees this as a big deal, and there are times when it definitely doesn’t matter. But, it’s typically best to create all of your ads at the start, rather than doing it later on and having to roll the dice on messing with results.

2. Use Text Variations

This feature has also been named Multiple Text Options or Multiple Text Optimization in the past. No matter what you call it, the functionality is the same.

When assembling your ad, you can create up to five variations of your primary text, headline, and description.

Text Variations

This is a great way to create variations while using only one ad. Meta will show combinations of text to people based on what they’re more likely to respond to. That could be due to what other people respond to, what the individual user has responded to in the past, the placement, and more.

Meta also generates primary text options that you can choose from using AI.

AI-generated Text Variations

I’ve found these rarely match up with my voice, so I don’t use them. But, it’s something worth testing out.

If you require control, you will not like this feature. There is no way of dictating how much a text variation is used — or whether it’s used at all. And since all of the variations contribute to the same ad, you won’t be able to see which combination led to the best results.

What you can do, though, is use the Breakdown by Dynamic Creative Element.

Breakdown by Dynamic Creative Element

A separate row will be generated for each variation, but you won’t see which combination performed the best.

Today’s advertiser needs to be okay with not always being in control while putting a certain amount of trust in the algorithm. This is a feature I regularly use, and I’m not overly concerned about “finding a winner.” Instead, I use it knowing that if I give the algorithm more options, I give it more opportunities to get the best possible results.

3. Use Dynamic Creative

Dynamic Creative was discontinued in June of 2024. Meta recommends using Flexible Ad Format instead. Read about the details of this update here.

Dynamic Creative is not a new feature (I first wrote about it in 2017), but it’s still useful.

Dynamic Creative combines multiple images, videos, and other ad components (primary text, description, headline, and CTA button) to find the best possible results while creating only one ad. This is similar to the Text Variations option, but it also includes creative and CTA buttons.

This feature is turned on in the ad set.

Dynamic Creative

When using “Single Image or Video,” you can upload a combination of up to 10 images and videos. It could be all images, all videos, or a combination thereof.

Dynamic Creative

You have the option of turning on Optimize Creative for Each Person. When this is on, ad creative and destinations vary depending on what an individual person may respond to.

Dynamic Creative

You can also test various CTA button options.

Like every option so far, this is not a true split test. If you’re hoping to test specific options against one another, this is not the option you want to use. Dynamic Creative is best for situations where you have several creative options, but you’re willing to give up control to the ad delivery algorithm.

As is the case with Text Variations, you will not see which combination of creative, text, and CTA button performs best. But, you can use Breakdowns to see how each individual item performed.

4. Run an A/B Test

A true A/B test is ideal for the control freak who has something very specific that needs to be tested. You want to find the best performer between two or more ads, free of overlap.

While you can run the options above indefinitely, an A/B test is meant to be temporary. You find a winner so that you can leverage it and turn off the losing variation. That’s why you’ll also need the benefit of time to run an A/B test.

Finally, keep in mind that your results are unlikely to be ideal during an A/B test. Your ads won’t be distributed optimally during this test because the entire goal is to segment your audience so that one half sees one version while the other half sees the other.

If you want to create a variation of an existing ad to test against the original, select the existing ad and click “Duplicate.” Then select “New A/B Test.”

A/B Test

For the variable that you want to test, select “Creative.” Then select the ad that you want to copy.

A/B Test

Pick the key metric that will determine a winner.

A/B Test

Then set a start and end date for the test. You can choose to have the test end early if a winner is found before the end date.

Unlike the other options listed in this post, an A/B test will give you a true winner — assuming that a winner is found and is statistically significant.

Which is Best for You?

The option that you choose for testing creative depends on your situation and what is important to you.

If you desire control and certainty and want to determine which ad is the top performer, use the A/B test option.

Otherwise, you’ll run a combination of the other three options. I rarely have a deep desire to know which ad is the top performer with an A/B test. It suggests that I already found two ads with preferred combinations of text and creative. And that is almost never the case.

When I create an ad set, I typically use multiple ads. Each ad will utilize a different format (video, image, or carousel), or maybe a different version of one of those formats. And each ad utilizes Text Variations.

Admittedly, I haven’t used Dynamic Creative for several years. But, I have heard that some advertisers still swear by it, and it’s not all that different from using the Text Variations optimization.

Like everything else, know your needs and style. Do what works for you.

Your Turn

Which approach do you take to creative testing?

Let me know in the comments below!