Experiments can be found in shortcuts or in the menu.
Experiments can be created on 2 or more existing variables, or new ones can be created. For example, a new campaign can be created with a different goal to see which performs better.
The first experiment can be done at the campaign or campaign group. There are several variables that can be tested here: image, video, ad text, age or gender, audience, or campaign optimization.
Whatever of what we want to test, it is recommended to go with a single variable, so that it is as relevant as possible. Up to 5 campaigns can be tested in parallel.
The winning version can be chosen according to several metrics, whether it is a cost per purchase or a standard event. The chosen metric does not influence in any way the experiment, it uses as a reference for our objective.
It is recommended that the experiment last a sufficient number of days, at least a week so that the system should have time to make a relevant comparison
Facebook recommends that the test power be over 80% for the experiment to be successful.
At the level of adsets, A / B Test experiments can be performed. We can choose up to 5 adsets to compare, the same number as for campaigns. The relevance of the test is as accurate as possible as we have only one differentiator within them. The relevance of the test is as accurate as possible as we have only one differentiator within them.
Holdout test can be used at the level of campaigns, campaign group, or account. The test involves conducting a campaign comparing Facebook users who had the opportunity to see our ad versus a control group that does not see the commercial. Then, it is calculated the incremental level contribution brought by those who saw the ads depending on what metrics we want. For example, conversions, but can be standard or custom events.
The brand survey determines the impact of Facebook ads on the brand and can be done at the level of account, campaign, or group of campaigns.
These are predefined questions and can be a maximum of 3, and the test is required to run for at least 14 days.No investment is required for this type of test, but the cost at the account level must be higher than RON 33,000 during the test.
The test results can be found in the Results section. A report is generated for each A / B test.
We recommend using Google Analytics for a more relevant analysis of results by using ad-level UTMs.