What is the role of A/B testing in Facebook ads

The social network Facebook has become a household name for its approximately 2.5 billion users. For marketers, however, Facebook is not just a platform for entertainment, but a workday, and above all, an invariable part of strategic planning.

It’s no secret that the apparently free-to-consumer network actually rakes in cosmic profits from ads. Businesses are competing with each other to reach its users. And since the ad revenue is really good, the Facebook team is working on helping businesses as much as possible to reach their target audience. Helping them actually helps your own profits. In short, a win-win situation.


Planning a Facebook ad campaign and not sure which look would get more audience engagement? Wondering which copy will score better? One of Facebook’s tools for successful ads is A/B Testing (Split Test). It gives you the opportunity to test different variations of your ad to assess which would be more effective – reaching more users at a lower cost or leading to more conversions. For this purpose, you have the opportunity to compare two or more options that differ by your chosen variable.

If you’re a restaurant that offers a lunch menu, for example, you can test whether users will respond more actively to a photo of lasagna or to a photo of moussaka, all else being equal.

The tool is valuable in that it allows you to test from the total budget and, already with the winning ad, launch the actual campaign with the majority of the budget. The result: A/B Testing improves the effectiveness of the advertising campaign, respectively the return on investment (ROI).

Where will you find it?

What variables can you test?

  • Creative – ie. you can experiment with the results that different creative elements would lead to – image design, video, ad text, headline, action button, style, etc. In other words, if you have a pet store, you can see if more users will click on the ad with a picture of a kitten, or if the image of a puppy will drive more traffic to your site.
  • Audience – If you’ve already determined the ad creative, you can test which audience responds better. Here you have the opportunity to check the potential customers from which city/area they respond to a greater extent, which age group, etc. For example, are women between 20 and 30 or rather ladies in the age group of 30-40 interested in a sale of hit book titles. See how to define your target audience.
  • Ad positioning – whether users will see the ad in their Facebook/Instagram news feed, whether it will appear within videos, whether via partner sites, in the right sidebar of Facebook, in story format, messages… the test options and there are many here. With the help of a test, you can find out where your users are more active – on Facebook or on Instagram.
  • Budget optimization – as you probably noticed – Facebook provides the option to optimize the budget and accordingly the reproduction of the campaign. You can test if and how it works for you.

You also have the opportunity to test more than one variable – in other words – see how two completely different strategies perform for your potential customers – different design with different text, or different text with different targeting. This will make it easier for you to decide which strategy is worth investing in. However, it is important not to test with too many variables at once. Otherwise, you risk the certainty of the advertising results.

Which variable to test?

Which variable you run the test with depends on which element would have the most impact on your campaign’s success rate. There is no universal formula. It all depends on the specifics of your business, the goals of the campaign and what your previous experience indicates. Unless you have a very high budget, it is pointless to run too many tests.

Good practices

The Facebook team recommends using a large enough audience for the test results to be correct. In case you targeted too small a segment of people, the findings may not be representative. And one more recommendation from the social network – do not use an audience that you use for a test in another ongoing campaign.

Testing is usually resorted to when you are planning a substantial campaign and have a budget that allows you to set aside a portion of it for testing. The means by which you test, in and of themselves, must also be sufficient to allow representativeness of the results.

 The duration of the campaign also has a bearing on the credibility of the results. It is reasonable for it to last at least a few days. Facebook’s recommendation is duration yes is 4 days, with the absolute minimum being between 1 and 3 days and the maximum for a test being 30 days. Of course, these periods are relative and largely depend on the specifics of your business and the behavior of the specific target audience towards the given products/services.

Budget allocation

Once you have determined the criteria by which Ad A and Ad B will be delivered – target audience, location for the ad, etc. in the Split test budget & schedule menu, you should enter your requirements on how Facebook should split the budget you set. With a click, you define exactly how the division should be carried out. It can be equal (Even Split) or one of the ads has x% higher budget than the other (Weighted Split). In both cases, the percentages are allocated based on the total budget you set, whether you set it to be the amount to be spent on a daily basis or for the entire campaign.

What budget to test with?

As for how much of the total campaign budget to set aside for the test, there is no exact rule. Leading is the amount of total funds you have available. If you’re not sure, though (and you can afford it), you can follow Facebook’s suggestion. The platform offers a budget on a daily basis and on an entire campaign basis. Be careful about the setting you note, as the difference in total spent is substantial.

Another thing to keep in mind – usually Facebook’s offers are in its favor ie. sets pretty generous sample budgets, so use your best judgment to trust his recommendation. On the other hand, Zuckerberg’s algorithm also sets minimum budgets below which you cannot fall if you want your campaign to run within the time frame you set.

As already mentioned, it is necessary to set a financial framework adequate to the test period not only to start the campaign, but also to have statistically correct results. In the event that the funds are too little, it is possible that you will not be able to achieve a sufficiently definite result.

Can I make changes?

Once you’ve hit the publish button on your campaign, you have the option to make changes while it’s in the review or planning stage. Once it’s active, you can’t edit the variable you’re testing, but you have the option to make other edits. In other words, if you are testing which image will have a better engagement rate at a lower price, you will not be able to change the image, but you will be able to adjust the criteria related to the audience. You also have the option to change the duration of the campaign and also to stop it.

How does Facebook determine the winning ad?

The ad that performed better is determined on the basis of a lower cost per result achieved according to the goal you set – impressions, engagement, purchase, etc.

Once the winning variant is known, Facebook will mark it with an asterisk for convenience. We advise you to take a closer look at the test results before proceeding further. This can be done via the View Charts menu, which appears as soon as you hover over a completed split test.

Based on the performance of the campaign, the platform runs simulations to determine the percentage probability that if you run the same test again, you will have the same results.

Here’s what the Facebook team says about the percentages, and more specifically when the percentages indicate that there is no clear winner in the test:

  • on 2 options – if you have less than 65% probability of the same winner in a retest
  • on 3 options – if you have less than 40% probability of the same winner in a retest
  • on 4 options – if you have less than 35% probability of the same winner in a retest
  • on 5 options – if you have less than 30% probability of the same winner in a retest


If the percentage you get is 75 or more, the results are considered strong enough for you to move forward with a larger ad campaign using the winning ad. However, Facebook specifies that each variant of advertising passes independently through the distribution system. Therefore, there may be some variance in the results that is not due to the variables you are testing.

The rеsults - in your inbox

Don’t worry about missing the end of the test campaign. Facebook takes care of alerting you. At the end of the set test period, you receive a notification both in the Ads account and directly on the email associated with Business Manager. You will also find a link in your inbox that will take you directly to the completed campaign and the results. Even without clicking on the link, however, Facebook synthesizes the basic information in the content of the email, namely which ad performed better, what was the result, what was the variable you tested, how much did you spend on the test, was the amount evenly distributed.


Knowing the results of the test campany, you can direct the rest of the funds allocated to the campaign to this ad, which is already proven to give better results. Facebook will prompt you to launch your actual campaign with the winning ad as soon as it informs you of the results. One option is this. However, you have other options – to run a new test campaign in case the results are not convincing (try a longer time frame and/or a higher budget), or to continue testing to check the performance of other variables in the ad . You also have the option to define your target audience even more precisely, for example, or even use the conclusions you have drawn from the results to create a new ad.

 If you have a clearly defined winning ad, you can directly start your actual campaign with it. Although you have taken into account the test results and launched the campaign with the presumption that you will get the best possible results, we recommend that you directly monitor how the indicators are moving – users reached, engagement (engagement), and cost per result achieved. Note that even if you are not satisfied with the results on the first day of the campaign, it is a good idea to wait 24 hours before making any changes. It often takes a little time to spin up a campaign, gain momentum, and lower your CPA.

In summary – if you want to be sure of the success of your campaign, there is no need to wonder whether to run a split test. The question is how much and what to test for maximum results.

Now, roll up your sleeves and start testing – you wouldn’t want to spend your budget based on your sixth sense alone, would you? 

Similar articles

20 Basic SEO Principles

Search Engine Optimization or SEO (Search Engine Optimization) is an effort that anyone with a web page should invest in to make their site more