Split testing in TikTok Ads Manager

This guide is built from TikTok's Business Help Center, starting with its page About Split Testing and the pages linked in each section below. It covers what a TikTok split test does, what you can test, how to set one up, how much budget and time it needs, how to read the result and how to end a test. The general method behind A/B tests (a hypothesis, one variable, enough data) is in our website A/B testing guide, and the same feature on Facebook and Instagram is covered in A/B testing in Meta Ads Manager.

What a TikTok split test does

A split test runs two versions of an ad group (or, in newer setups, two campaigns) side by side and tells you which one did better. What makes it different from simply running two ad groups at once is how the audience is handled. TikTok says it will split your audience into two equal groups, with each group seeing only one ad group, so the two versions don't bid against each other for the same people. TikTok calls that problem "squeezing", and its help page lists preventing it as one of the main benefits.

At the end, TikTok's model compares the two on a key metric you picked and names a winner only if the gap is statistically significant. TikTok describes the test as designed to run an accurate A/B test with a 90% confidence rate. If you run two ordinary ad groups without the split test toggle, both can reach the same users and the comparison is less clean.

TikTok's May 2025 product post on campaign-level split testing announced a beta that compares two whole campaigns, including campaigns with different objectives (its example is Reach vs Traffic), and can generate a Smart+ campaign from a manual one to test the two against each other. It works the same way: the audience is divided randomly into groups A and B, and each group sees only its own version.

What you can test

TikTok's About Split Testing page, last updated January 2026, lists these test variables:

• Targeting: for example a broad audience against a custom audience.
• Placement
• Bidding and Optimization: for example two bid strategies or optimisation goals.
• Budget Strategy
• Creative Assets, Creative and Catalog
• Custom: campaign level tests for a combination of variables
• Smart+: a Smart+ campaign against a manual one.

The 2025 product post named five (Targeting, Bidding & Optimization, Creative, Smart+ and Multi-Variable), so the list you see depends on your account and objective. Whatever you pick, everything else in the two versions stays the same; that is what lets the result point to the one setting you changed. For the targeting options themselves, see our TikTok ads targeting guide, and for bid strategies, TikTok bidding strategies and budgets.

Setting up a split test

TikTok's help page describes the split test toggle as part of creating a new campaign, so plan a test as a new campaign rather than a change to a running one. Its How to create a split test page gives these steps:

1. In TikTok Ads Manager, open the Campaign tab and click Create.
2. Choose an advertising objective.
3. Turn on the Create a split test toggle and click Continue.
4. Complete the ad group settings. TikTok's note here: Set the test schedule dates to at least 7 days so the system has time to collect a useful sample.
5. Complete the ad settings. This first ad group becomes the control and will start on its scheduled date.
6. In the Split test tab, select the split test variable.
7. Select a Key metric. TikTok says The system uses the Key Metric to compare the two ad groups, so pick the result you actually care about, such as cost per purchase, not a metric that is just easy to move.
8. Set up the test ad group with the one difference, then click Complete.

If the test measures conversions, the TikTok Pixel and ideally the Events API need to be sending those events first. TikTok's own testing article asks for both. Our TikTok Pixel guide covers the setup.

Budget and testing power

A test with too little spend ends with "no winner" because neither side gets enough results. TikTok's split test best practices page says that when you create the test, the system will provide the Estimated Testing Power based on the budget you set. Power is the chance of detecting a real difference between the two ad groups if there is one. TikTok's advice: We recommend a budget that gives you a power value of at least 80%.

The same page has three more tips that affect how much data you get:

• Make the two versions clearly different. TikTok recommends that the settings differ significantly, because two near-identical versions tend to produce similar results and no winner.
• Use a large audience to avoid an insufficient sample size.
• Start from a hypothesis, so you know which variable to test and what you will do with the answer.

In a 2024 business blog post, How to test, learn, and scale your TikTok campaigns, TikTok gives a rule of thumb: set a budget of at least 20x CPA based on your campaign history. That is a planning number; the testing power estimate in Ads Manager is the figure to check for your own test.

How long to run it

TikTok's help pages and its blog give different minimums, so here they are side by side:

What	TikTok's number	Where it says so
Minimum test length	At least 7 days	Create a split test and best practices pages (help center)
Recommended test length	At least 14 days	2024 business blog post
Maximum test length	60 days	Best practices page
Confidence level for a winner	90%	About Split Testing and results pages
Recommended testing power	At least 80%	Best practices page
Budget rule of thumb	At least 20x CPA	2024 business blog post
Extra time after editing a running test	An additional 7 days	Edit a split test page

The best practices page explains the trade-off: Testing for less than 7 days may not be long enough to determine a winner, while long testing may lead to a shortfall in the budget. In our experience, seven days also helps because it covers each day of the week once.

Editing a running test

Some settings are locked. TikTok's edit a split test page says that certain settings like your testing variable and ad information can't be modified; other ad group settings can. Its advice is not to edit often, because the system needs time and budget to relearn, and if you change a control setting, make the same change in the other ad group so the comparison stays clean:

1. Find the first ad group in the split test, hover over the name and click Edit.
2. Make your edits and click Save.
3. Make those same exact edits to the other ad group.

After an edit, TikTok recommends giving the system an additional 7 days and extra budget to pick a winning ad group. The best practices page adds that edits may also send the ad group back into review. How TikTok's learning phase reacts to edits in general is covered in our bidding guide.

Reading the result

To see a test's result, per TikTok's split test results page:

1. In TikTok Ads Manager, click Campaign, then the Ad group tab.
2. Find the ad groups in the split test, hover over the name and click View Data.
3. Click the Split Test tab. You can also add custom columns for other metrics.

When the test finishes, TikTok judges the winner on the key metric you picked at setup and shows one of two results:

• Winning ad group found (90% confidence level)
• No winning ad group found (90% confidence level)

"No winner" is still an answer: at this budget and length, the variable you changed made no clear difference to that metric. That can mean the two versions really perform alike, or that the test was too small to tell. The testing power estimate you saw at setup helps you judge which. The other View Report tabs are explained in our TikTok ads reporting guide.

Ending a test early or keeping the winner

A split test ends on its scheduled date, but you can stop one side before that. TikTok's pause a split test page notes that It is best practice to wait for the system to select a winner before pausing one ad group, since stopping early leaves you without a result. The steps:

1. Go to the ad group within your campaign.
2. Click the On/Off toggle next to the ad group name.
3. In the pop-up, choose Pause This Group Only or Pause Both.

What happens next, per TikTok: the paused ad group becomes "inactive", the split test icon disappears, and The budget from the paused ad group will transfer to the active ad group. Audiences then only see the active one. Pausing the losing ad group is also how TikTok tells you to keep running the winner once one is found. If you reactivate a paused ad group, it runs as a normal ad group without the split test; to test again, create a new split test.

If you later copy a manual campaign into Smart+ with TikTok's Duplicate feature, the split test setting is removed, as noted in our Smart+ guide.

iOS 14.5+ dedicated campaigns

Split testing is limited for iOS 14.5+ dedicated app campaigns. TikTok's page on creating an iOS 14.5+ dedicated campaign says split tests are not available for dedicated campaigns unless you are beta-testing the feature, and where they are, they are available for up to 2 such campaigns, which count towards the account's 15-campaign quota. The results page adds that conversions for these campaigns may take up to 72 hours to be posted back to Ads Manager, so wait before reading the last days of a test.

How we run TikTok split tests

This part is our own practice, not TikTok's guidance:

• We write the hypothesis and the decision before launch: "if B wins on cost per purchase, we move the budget to B's setting".
• We test one big difference at a time (for example broad vs a custom audience, or two clearly different video concepts), not small tweaks.
• We pick the key metric closest to revenue that has enough volume, and set the budget so Ads Manager shows a testing power of at least 80%.
• We run at least 7 days, usually 14, and leave both ad groups untouched until the result is in.
• We keep a log of each test, its result and what changed afterwards, so the same question is not tested twice.

If you want help planning tests for your TikTok account, see our services.

Frequently asked questions

What is a split test on TikTok?

It is a feature in TikTok Ads Manager that runs two versions of an ad group or campaign against each other. TikTok splits the audience into two equal groups, each group sees only one version, and the system names a winner on your key metric if the difference reaches 90% confidence.

How long should a TikTok split test run?

TikTok's help pages say at least 7 days and at most 60 days. A 2024 TikTok business blog post recommends at least 14 days for more reliable results.

How much budget does a TikTok split test need?

Enough for Ads Manager's Estimated Testing Power to reach at least 80%, which TikTok recommends. Its 2024 blog post also gives a rule of thumb of at least 20 times your CPA from past campaigns.

Can I add a split test to an existing TikTok campaign?

TikTok's help page only describes the Create a split test toggle when you create a new campaign, so plan the test as a new campaign. Once a test runs, the testing variable and ad information can't be changed; other settings can, but should be changed in both ad groups.

What happens if I pause one ad group in a split test?

The paused ad group becomes inactive, the split test icon disappears, its budget moves to the active ad group and audiences see only the active one. TikTok recommends waiting for the system to pick a winner before doing this.

For testing on other platforms, see A/B testing in Meta Ads Manager and our website A/B testing guide. For more on TikTok, see TikTok ads basics and TikTok Spark Ads.