Split Testing Ads Creative

Published by Dina Taitelbaum on

are you a/b testing your ads creative?

Split Testing Ads Creative

Why Test Ads Creative?

Too many marketers out there throw in a few creatives ideas that seem appealing to them, combine them with their main ESPs & USPs texts, and launch them as new ads. Some may experiment enough to mix and match the creatives with the texts.

Be honest: when was the last time you split tested your ad creatives? What about the elements within your creatives?

That’s because creating original content all the time, that speaks your brand’s language, represents the look & feel while also drives sales takes a lot of time and thought. In fact, it takes time, thought, planning, design skills, information on past campaigns, analysis of those campaigns, and finally, lots of patience to put all that information together and execute a new creative with compelling copy variations.

Ask any marketing influencer, they will all speak of the importance of split testing ads to figure out what content speaks to your target audience.

Split testing ads is highly important for any company and brand because it helps you understand which value proposition and image/video combination drive better results.

Which keywords touch the right “nerves” of your consumers?

Which elements like colors, frame speed, volume, text style, concept, amount of people, dimension, etc (the list of creative elements is endless, as you can imagine) grab their attention?

Creative plays a huge role in the performance of your ads. In fact, a Nielsen study shows that creative impacts 47% of the performance of the ads – that’s larger than any other aspect of advertising. Creative mistakes can cause a significant amount of media budget loss so spending some time and money split testing to figure out what works will be really worth it in the long run. 

Why Split Testing Ad Creative is a Challenge

Running A/B tests (aka split tests) is as challenging as it is important. 

Accurate planning and organization are crucial for identifying which creative variations you’re testing. Thus, make sure you label your ads (tests) accurately so you can mix and match your variants while keeping tracking of them. This will be highly important for the later analysis.

The analysis of your split tests is just as crucial as its execution, if not more. When your ads have run for about a week, you will be able to figure out which variants performed the best. Additionally, we recommend including the ad’s goals and funnel stages within the description. Here’s what these variations could look like:

  • image-donut-yellow-satisfyslogan-25off%-awareness-comment
  • image-donut-yellow-satisfyslogan-nodiscount-awareness-comment
  • image-donut-yellow-satisfyslogan-25%-conversion-linkclick
  • image-donut-yellow-satisfyslogan-nodiscount-conversion-linkclick

In the example above, we’re checking if the discount offer works better in the awareness campaign or in a conversion one. Note that nothing else is changing in the creative (on purpose!) because the moment you change any other aspect, your test is no longer controlled and you can’t tell what impacted the result.

split testing ads example with donuts

This example tests only one type of creative element but hopefully, this gives you an idea of how to organize your test so you can identify exactly what you’re testing and analyze it at the end.

Here’re a couple of more testing options, just for fun (but notice that only 1 element is changed each time):

split testing ads example of copy
ab testing discount offer on ads

BTW: these would be called…

  • image-donut-yellow-satisfytext-nodiscount-awareness-comment
  • image-donut-yellow-ordertext-nodiscount-awareness-comment
  • image-donut-pink-satisfytext-nodiscount-awareness-comment
  • image-donut-pink-ordertext-nodiscount-awareness-comment
  • image-donut-pink-ordertext-25%-awareness-comment

Testing many variables at once, not as controlled experiment

A less controlled experiment when your split test has changes of multiple variables is a much harder test to analyze.

However, it would be a more affordable solution because you’d be running a lot fewer ads. Practically, most marketers do that simply because it’s easier to implement (but harder to understand).

We’d be happy to hear how you’ve been split testing your ads (if you’re not a Pudding.ai user, of course), what is your method? How do you analyse your tests and draw conclusions on what creative elements help the performance?

Reach out with your response to dina@pudding.ai and we’d love to include your method in our next article about split testing ads.

What Creative Ad Elements Should You Be Split Testing?

  • Headlines
  • Images, videos
  • CTAs, offers
  • Ad copy, its length, emojis, keywords
  • Text on image and its style
  • Colors, brightness 
  • People presence
  • Frame speed and count (video)
  • Music from the start (video)
  • Subtitles (video)
  • Dimensions
  • Number of objects
  • Concept 
ads creative report CTA

A/B Testing Publishing Strategy

As you may have already thought to yourself, testing all these elements and variants means paying for each one of them and how much is this going to cost? There’s nothing you can do about it. Split testing ads is not a cheap business but in the end, you figure out what actually works for your business and audience. 

In fact, besides split testing ads creative there needs to be a publishing strategy as well. Here’re additional aspects that can impact your creative performance:

  • Audience
  • Time of day
  • Channels

This means the same “winning” creative may not work for a different audience, at a different time of the day, or on another channel!

Ideally, to conduct a thorough a/b test for your ad creatives the campaigns should also be testing your publishing strategy. 

Dos and Don’ts of Split Testing Ads


  • Test multiple value propositions
  • Keep in mind the campaign’s funnel
  • Let your campaigns run for at least a week
  • Change only one creative element (including value propositions) for each test variant
  • Test your publishing strategy


  • Assume the best selling point
  • Forget to differentiate your ads for different stages of your customer journey aka funnels
  • Judge your results based on the first couple of days
  • Forget to test out different audiences
  • Try various channels, like Google, TikTok, YouTube, etc. (this also relates to testing different audiences)

For additional tips, check out our #PuddingReport blog posts. Meanwhile, go ahead and drop your Do’s and Don’ts in the comments section below.

The Future of AI and A/B Ad Testing

AI technology is becoming more and more valuable in the advertising world. It’s becoming increasingly challenging for marketers to juggle multiple ad platforms in an effort to reach various audiences.

Artificial intelligence is becoming a much-needed solution to measure efforts, create content, publish, analyze, chat, and much much more. In 2019, the machine learning application industry received $37 billion of funding in the US.

The whole online marketing process is being impacted by AI! According to Semrush, marketing and sales departments prioritize AI technology and machine learning for their success more than any other department (40%).

AI tech is definitely on the rise while a/b testing ads… seems to be slowly decreasing in popularity.

a/b testing keyword overview screenshot from ubersuggest
Keyword overview from Niel Patel’s Ubersuggest SEO Tool
a/b testing ads keyword overview screenshot from ubersuggest
Keyword overview from Niel Patel’s Ubersuggest SEO Tool

It’s not surprising that with the technological advancement, a tedious practice of a/b testing ad variations would be replaced. Isn’t it the purpose of technology? To simplify those tedious tasks?

We definitely believe so. Leave split testing ads in 2019!

How Pudding.ai Can Help

Pudding.ai is a creative analysis solution that helps marketers understand which elements of their ad creative help or hurt the performance. This means it eliminates the need for split testing ads altogether.

With Pudding, the performance of all elements within your creatives (colors, text and its positions, human presence or lackof, frame-speed, voice-over, and countless more) are measured against your chosen KPI.

This means, all you have to do is pick your goal and view which creative elements help to reach it. The AI knows to identify what works, what doesn’t, why and it gives you insights and suggestions on what you can do to improve your performance.

pudding.ai ads creative insights

Pudding.ai helps marketers avoid all the dreadful work related to split testing ads. All the creative element variants and their labeling is done completely automatically by Pudding.ai. For those who include unique branding elements within the creative, there is also a way to teach the AI those unique labels. Then, in all future analyses the AI will include that label and it’s impact to the ads creative performance.

Want to audit your ads creative and see which elements impact your performance? Go ahead and signup for a free report and we’ll send you your ads creative insights directily in your mailbox!

schedule a demo blog cta