A/B testing is the only way to know what works instead of guessing. On TikTok, the two axes that matter most are creative and audience.
Rule number one: test one variable at a time. If two ad groups differ in both video and audience, you cannot tell which factor drove the difference. Keep everything identical except the one thing being compared.
Creative tests: same audience, same budget, different videos. Test the hook first (the first 3 seconds drive most of CTR), then format (real people vs graphics), then CTA. Audience tests: same creative, different audiences — for example broad vs interest, or two age brackets.
On sample size: wait until each variant reaches at least 1,000–2,000 impressions for CTR goals, or 30–50 conversions for conversion goals, before concluding. Close the test after 3–7 days, keep the winner, and iterate on the next variable.
Log every round's results — within a few months you will have your own playbook of what works for your customers, an asset worth more than any generic tips.