TikTok split test: how to A/B test your ads

A split test (an A/B test) keeps everything the same except one variable, splits your audience into two equal groups, and shows each group only one ad group. That stops the two versions competing for the same people, which TikTok calls squeezing.

TikTok says the test runs at a 90% confidence rate, and the system only picks a winning ad group when the result is statistically significant.

• Targeting
• Placement
• Bidding and optimisation
• Budget strategy
• Creative assets
• Catalogue
• Creative
• Custom: a campaign-level test of a combination of variables
• Smart+

Source: TikTok Business Help Center: About Split Testing, checked 15 October 2026

The split test is a toggle you turn on while creating a campaign. TikTok's steps:

1. Go to TikTok Ads Manager and click the Campaign tab.
2. Click Create and choose an advertising objective.
3. Turn on the Create a split test toggle and click Continue.
4. Complete the Ad group tab. Set the schedule to at least 7 days so the system can collect a useful sample.
5. Complete the Ad tab. This is your control ad group.
6. In the Split test tab, select the split test variable.
7. Select a key metric. The system uses it to compare the two ad groups and pick the better one.
8. Set up your test ad group and click Complete.

Source: TikTok Business Help Center: How to create a split test, checked 15 October 2026

Too little time or budget is the usual reason a test can't find a winner. TikTok's guidance:

• Run it for at least 7 days. The maximum is 30 days.
• Under 7 days may not be long enough to find a winner. A very long test can run short of budget.
• When you create the test, Ads Manager shows an Estimated Testing Power based on your budget. TikTok recommends a budget that gives a power value of at least 80%.
• Power value is the likelihood of detecting a difference between the two ad groups.
• Use a large audience so the sample isn't too small.

Source: TikTok Business Help Center: Split Test Best Practices, checked 15 October 2026

TikTok's best practices, in order:

1. Write down a hypothesis first: what you're testing and how the answer will change your ads.
2. Make the two versions clearly different. If they're too similar, both produce similar results and the system can't pick a winner.
3. Set a budget that reaches at least 80% estimated power.
4. Leave it alone once it's running. Changes can affect the result or send the ad group back into review.

Source: TikTok Business Help Center: Split Test Best Practices, checked 15 October 2026

At the end of the test you can choose to keep running the winning campaign or ad group with one click. You can also pause an ad group yourself, or let the system pause it once it picks a winner.

TikTok suggests running further split tests to keep refining what works, so treat each result as one answer, not a final rule.

Sources: TikTok Business Help Center: About Split Testing, checked 15 October 2026; TikTok Business Help Center: Split Test Best Practices, checked 15 October 2026

Check these first:

• The two versions were too similar to produce a measurable difference.
• The budget gave a low estimated power value. Aim for 80% or more.
• The test ran for less than 7 days.
• The audience was too narrow, so each half got too few results.
• Settings were edited mid-test, which can skew the result or trigger another review.

Sources: TikTok Business Help Center: Split Test Best Practices, checked 15 October 2026; TikTok Business Help Center: About Split Testing, checked 15 October 2026

Zolyp, which publishes this guide, is an AI social media manager: it plans, writes, designs and schedules posts for Instagram, Facebook, LinkedIn, Google Business, X, TikTok, YouTube, Pinterest, and Threads, and you approve the week before anything goes out. It doesn't run paid tests. Its organic posts are a cheap way to see which hooks and ideas people respond to before you put ad budget behind one.