Split testing lets you test different versions of your ads so you can see what works best and improve future campaigns. For example, you can test the same ad on two different audiences to see which ad performs better. Or, to test two delivery optimisations to determine which selection yields better results.

To get started, navigate to Ads Manager and create a split test. Use this guide to understand the basics of split testing, including variables, budget and scheduling.

How split testing works

Facebook's split testing feature allows advertisers to create multiple ad sets and test them against each other to see what strategies produce the best results. Here's how it works:

• Split testing divides your audience into random, non-overlapping groups.
• This randomisation helps to ensure the test is conducted fairly because other factors won't skew the results of the group comparison. It also ensures that each ad set is given an equal chance in the auction.
• Each ad set tested has one distinct difference, called a variable. Facebook will duplicate your ads and only change the variable that you choose.
• To get the most accurate results from your split test, you'll only have the opportunity to test one variable at a time. For example, if you test two different audiences against each other, you can't also test two delivery optimisations simultaneously because you wouldn't know for sure which change affected the performance.
• Split testing is based on people, not cookies, and gathers results across multiple devices.
• The cost per result of each ad set is calculated and compared. The ad set with the lowest cost per result, such as cost per website purchase, wins. We make these calculations with Facebook's attribution system. We use data from the test itself and thousands of simulations based on it, which helps us determine our confidence level in the results.
• Once the test is complete, you'll receive a notification and email containing results. These insights can then fuel your advertising strategy and help you design your next campaign.

Objectives available for split testing

Facebook split testing supports the following business objectives:

• Traffic
• App Installs
• Lead Generation
• Conversions
• Video Views
• Catalogue Sales
• Reach
• Engagement
• Messages
• Brand Awareness

Variables available for split testing

Advertisers will have the option to test one of the following variables. You can test five different strategies with one of these variables.

• Target audience
• Delivery optimisation
• Placements
• Creative
• Product sets

Learn more about the variables that you can test.

Setting a budget and schedule

Your split test should have a budget that will produce enough results to confidently determine a winning strategy. You can use the suggested budget that we provide if you're not sure about an ideal budget. (We calculate the suggested budget by analysing successful split tests that have run in the past and that had settings similar to your test). We'll also provide a mandatory minimum budget to help guide you. The budget and audience will then be divided between the ad sets. You can choose to divide it evenly or weight one more than the other(s), depending on your preference.

We recommend 4-day tests for the most reliable results, and if you aren't sure about an ideal time frame, you can start with four days. In general, your test should run for at least three days and no longer than 14 days. Tests shorter than 3 days may produce insufficient data to confidently determine a winner and tests longer than 14 days may not be an efficient use of budget, as a test winner can usually be determined in less than 14 days.

For this reason, we recommend a test between 3-14 days for tests created in the API. When creating a split test in Ads Manager, you must create a test with a schedule between 3-14 days.

Next steps

When the test is over, you'll receive a notification in Ads Manager and an email with the results. Learn more about how the winning ad set is determined. Once you receive your split test results, you can review them to discover insights about the best-performing ad set. These insights can help you determine your advertising strategy and design your next campaign.

A split test lets you test different versions of your ads so you can see what works best and improve future campaigns.

In Ads Manager, there are three ways to create split test:

• Guided creation. We guide you through the process of creating a split test, and your ads will be ready to run once you've completed the workflow. You can choose this method if you're a new advertiser or if you're more comfortable with a step-by-step process.
• Quick creation. You create the structure for a split test that can be finalised later.
• Duplication. Ad a new ad set or ad to turn an existing campaign into a split test campaign.

Learn about best practices for split testing.

A split test lets you compare different versions of your ads so you can see which performs better. With each test, you can use one of the following variables:

• Creative
• Audience
• Optimisation event
• Placements
• Product set

Testing creative

Testing ads with different creative helps you understand what images, videos, text, headlines and/or calls to action (CTA) perform better.

Strong creative assets can help your ads stand out, and you can use creative testing to understand what your audience is more likely to engage with.

To get started, create a split test and select Creative as your variable. You'll then be able to create between two and five different versions of your ad. If you want to be able to attribute your results to a single creative element, we recommend only testing one (e.g. different images for each ad) and keep everything else identical.

Example tests to run:

• Ad with a video compared to ad with single image
• Ad with one video compared to ad with different video compared to ad with another different video
• Ad with one headline compared to ad with different headline compared to ad with another different headline

Note: You can still test ads with multiple differences to see which one performs better.

Testing audiences

Testing different audiences helps you understand what types of people are more likely to respond to your ads.

To get started, create a split test and select Audience as your variable.

Example tests to run:

• Women, aged 21-30 compared to women, aged 31-40 compared to women, aged 41-50
• People living in London compared to people living in Paris compared to people living in New York City
• A Custom Audience of people who have recently visited your website compared to people who may want to buy your products based on interest targeting

Testing optimisation events

How well your ad performs can depend on what you optimise delivery for. For example, your ad may perform better if you optimise for landing page views instead of link clicks. Test different optimisation events to see which one leads to better results.

To get started, create a split test and select Delivery optimisation as your variable.

Example tests to run:

• Ad set optimised for conversions compared to ad set optimised for link clicks
• Ad set optimised for landing page views compared to ad set optimised for link clicks compared to ad set optimised for impressions
• Ad set optimised for conversions with a 1-day post-click conversion window compared to ad set optimised for conversions with a 7-day post click conversion window

Testing placements

Testing placements can help you understand what platforms (e.g. Instagram, Facebook) are most effective for your ads, and where it's best to show them on each platform (e.g. Stories, feed).

To get started, create a split test and select Placements as your variable.

Example tests to run:

• Automatic placements compared to customised placements
• Mobile placements compared to desktop placements

Note: Limiting placements can negatively impact ad delivery. After you run a split test and understand which placements lead to better performance, you can use those information results to inform future decisions. However, in general, remember that it's recommended to have as many placements as possible.

Testing different product sets

Testing ad sets with different product sets helps you understand which product set in a Facebook catalogue performs better in a catalogue sales campaign. A product set is a group of items from your inventory that you designate in your catalogue. You can use a product set to control the items that appear in your ad.

To get started, create a split test using the Catalogue Sales objective and select Product set as your variable. You'll then be able to create between 2 and 5 different versions of your ad set.

Note: The audience for each product set may be different. Because people automatically see products based on their behaviour and relevance, your audience targeting doesn't overlap when you create a product set split test.

Example tests to run:

• Product set with items that are less than GBP 50 compared to product set with items that are more than GBP 50
• Product set with shoes from one brand compared to product set with shoes from a different brand compared to product set with shoes from another different brand
• Product set with all brands from your catalogue compared to product set with only private-label brands from your catalogue

Split testing compares different versions of your ads so you can see what works best and improve future campaigns.

Use these best practices to create a split test with clearer and more conclusive results.

Test only one variable for more conclusive results

You'll have more conclusive results for your test if your ad sets are identical except for the variable that you're testing.

Testing a single variable is important when you're experimenting with different creatives. If your ad sets have several creative aspects that vary (e.g. different images, different headlines and different text), then when the winning ad set is declared, you won't know which factor to credit. To make this easier to set up, after you create your first ad set, we automatically duplicate its settings for your other ad sets (aside from the variable you're testing, which should be different).

Bear in mind that testing multiple creative variables can still yield valuable results and learnings for future campaigns.

Focus on a measurable hypothesis

Once you figure out what you want to test or what question you want to answer, create a testable hypothesis that enables you to improve future campaigns. For example, you might start with a general question such as "Do I get better results when I change my delivery optimisation?" This can be refined to something more specific, such as "Do I get a lower cost per result when I optimise for link clicks or landing page views?" From there, you can set an specific hypothesis, such as "My cost per result will be lower when I optimise for landing page views." This helps you interpret your results and empowers you to take specific action on future campaigns.

Use an ideal audience for the test

Your audience should be large enough to support your test, and you shouldn't use this audience for any other Facebook campaign that you're running at the same time.

Overlapping audiences may result in delivery problems and contaminate test results.

Use an ideal time frame

We recommend 4-day tests for the most reliable results, and if you aren't sure about an ideal time frame, you can start with four days.

In general, your test should run for at least three days and no longer than 14 days. Tests shorter than three days may not produce enough data to confidently determine a winner, and tests longer than 14 days may not be an efficient use of budget, as a test winner can usually be determined in 14 days or fewer.

For this reason, we recommend a test between 3-14 days for tests created in the API. When creating a split test in Ads Manager, you must create a test with a schedule between 3-14 days.

Your ideal time frame (within the 3 to 14-day time period) may depend on your objective and business vertical. For example, let's say you're running a split test to determine which delivery optimisation leads to a lower cost per result. You know that people tend to take longer than four days to convert after seeing an ad, so you'd want to run your test for more than four days.

Set an ideal budget for your test

Your split test should have a budget that will produce enough results to confidently determine a winning strategy. You can use the suggested budget that we provide if you're not sure about an ideal budget.

Learn more about the basics of Facebook split testing or learn about creating your first split test.

Selecting a budget:

Your split test should have a budget that will produce enough results to confidently determine a winning strategy. You can use the suggested budget that we provide if you're not sure about an ideal budget. We calculate the suggested budget by analysing successful split tests that have run in the past and that had settings similar to your test. We will also provide a minimum budget, and this is the lowest possible budget that you can select for your campaign.

Once you select a budget, the budget and audience will then be divided between the ad sets. You can choose to divide it evenly or weight one more than the other(s), depending on your preference.

Selecting a schedule:

We recommend 4-day tests for the most reliable results, and if you aren't sure about an ideal time frame, you can start with four days. In general, your test should run for at least three days and no longer than 14 days. Tests shorter than three days may produce inconclusive results and no winning ad sets, and tests longer than 14 days may not be an efficient use of budget since a test winner can usually be determined in 14 days or sooner.

For this reason, when creating a split test in Ads Manager, you must create a test with a schedule between three and 14 days. We recommend a test between three and 14 days for tests created in the API.

Learn more about split testing and how you can use split test to improve your ads' performance.

Once you've created a split test, you might want to make changes to your campaign, ad set or ad. For example, if you notice that the headline in your ad has a typo, you'll want to fix the ad without cancelling the whole test. To do that, just edit your ad set or ad in Ads Manager just as you would in any other situation. Learn how.

However, remember that changing your split test after it has started may impact its results. Here are some best practices for editing split tests:

• Only edit your campaign, ad set and ad when it's absolutely necessary. All edits can affect the accuracy of your results, so you should only edit your campaigns when it's necessary. If the change isn't necessary, you can always create a new test.
• Avoid making significant changes to the variable that you're testing. If you make changes to the variable while the test is still running, you may not be able to know why the winner won.
• Use caution when making changes to things other than your variable. Editing the audience, placement, creative and other factors can affect the results and undermine your understanding of what drove the results. For example, if you test two different audiences with the same ad creative, and then later change the ad creative, this change might cause one audience to interact more with the ad. When the split test is done, you won't know if the winner was determined based on the audience or the ad creative.

You may also want to cancel your test. To do that:

1. Go to Ads Manager.
2. Tick the box next to the campaign that includes the split test you want to cancel and open the side panel by clicking the icon.
3. From the drop-down menu links, click Cancel test.
4. Click Confirm and close.

After cancelling your split test, your budget and audience will no longer be divided between your ad sets. If you cancel your split test, no winner will be determined and you won't get a notification or email with results.

The winning ad set is determined by comparing the cost per result of each ad set. The ad set with the lowest cost per result, such as cost per website purchase, is calculated in Facebook's attribution system.

Based on the data from the test, Facebook simulates the performance of each variable tens of thousands of times to determine how often the winning outcome would have won.

Once a winner is determined, you'll receive a notification in Ads Manager and an email will be sent to you with the results.

Learn more: About split testing

A split test lets you test different versions of your ads so you can see what works best. By analysing your split test results, you can determine what changes you might want to make to future campaigns.

Viewing your results

Ads Manager results

In Ads Manager, you can:

• Check the winning ad set or ad in the reporting table after your test is over. It will have a star icon next to it. Learn more about how winning ad sets are determined.
• Apply a filter so that only campaigns, ad sets and ads that are part of split tests will be shown in the reporting table. To do this, click Filter and choose Split test from the menu.
• Learn more about your results and the winning ad set by opening the reporting panel. Select your split test campaign by ticking the box next to the campaign and clicking View Charts or the icon . The reporting side panel will open and show the:

  • Winning ad set
  • Cost per result for each ad set (or ad for creative tests)
  • Confidence level, which tells you how likely it is that we'd get the same winner if you ran this test again

Email results

When your split test is complete, you'll receive an email with its results, which include:

• The winning ad set and how likely it is that we'd get the same winner if you ran the test again
• Your split test settings including:

  • The variable you tested
  • Whether your split was weighted or even
  • Budget
  • Schedule
• A breakdown of:

  • Results. The number of times your ad set or ad got the result associated with your campaign objective
  • Cost. The average cost per result from your ads
  • Amount spent. The total amount of money you've spent on your campaign, ad set or ad during its schedule
• How your budget was distributed across the ad sets or ads you tested
• A link to view the full campaign in Ads Manager

Understanding your results

The winning ad set is determined by comparing the cost per result of each ad set or ad based on your campaign objective. We also provide a confidence level, which represents how likely it is that we'd get the same results if you ran the test again.

For example, let's say that you run a creative test with one video ad, one single image ad and one carousel ad. We determine that the video ad was the winner with the lowest cost per result and a 95% chance that you would get these results again. With these results, we recommend that you adopt the winning strategy.

Use this guide to interpret results with confidence levels below 75%, broken down by the number of ad sets or ads you tested. If your results are less confident than the percentages listed below, the variables you tested are not likely to be a key factor in the performance of your ad, relative to each other:

• Two ad sets or ads tested. Less 65% confidence means that the variables you tested performed similarly.
• Three ad sets or ads tested. Less than 40% confidence means that the variables you tested performed similarly.
• Four ad sets or ads. Less than 35% confidence means that the variables you tested performed similarly.
• Five ad sets or ads. Less than 30% confidence means that the variables you tested performed similarly.

If your test has a confidence level above these numbers, you can use it going forwards. Regardless of the number of ads or ad sets you tested, a percentage above 75 indicates that we consider the results very clear and actionable. This means that we recommend using what you learn from the test when making decisions about relevant campaigns in the future.

Note: Every ad set in a split test goes through the delivery system independently, which means there may be a degree of variance between the results for your ad sets caused by the delivery system and not solely by ad set differences.

Next steps: Improving future campaigns

• Create a new ad from the winning ad set directly from the results email
• Reactivate the winning ad set in Ads Manager
• Create a new campaign based on what you learned

You could also run more tests to refine your strategy further. For example, if an aged 18-30 audience outperformed an aged 31-43 audience, you could narrow that further to see how that affects performance. You might next test an aged 18-30 audience against an aged 24-30 audience. These sorts of refinements are important for getting the most out of your campaigns.

If your results have low confidence, you can test the campaign again with a longer schedule or higher budget. More time and budget typically produces more data, which helps us be more confident in your results.

A split test lets you try different versions of your ads so that you can understand which version resonates more with your audience or drives better results.

Once your split test is complete, you'll receive an email with results and can review these in the email or in Ads Manager. If you have questions about your results, you can read our guide on understanding your results.

After you've reviewed your results, you may find that no winner was declared or that your ads under-delivered. In that case, you can use the tips below to troubleshoot your split tests and improve future tests.

If you want to run more tests, we've provided some test examples below that may help you refine your advertising strategy or improve your ad performance.

Learn more about the basics of setting up a split test.