> Quick answer: Start broad. Enable Smart Targeting. Build custom audiences from real data. Test for 7+ days. Aim for 10-20% CPA improvement per quarter.
Understanding TikTok's Targeting Approach
TikTok's ad system rewards advertisers who give the algorithm room to work. Tight restrictions hurt performance more than they help.
Why broad targeting outperforms narrow restrictions
Per TikTok's Business Help Center, broad targeting (80%+ audience reach) achieves 15% lower CPA and 20% higher conversion rate compared to narrow targeting. That's a big gap. Narrow restrictions cut off the signal TikTok's machine learning needs to find buyers efficiently.
Most advertisers over-restrict. They stack interests, behaviors, and demographics all at once. The result is a small audience the algorithm can't optimize across. Start open. Layer restrictions only when data tells you to.
How TikTok's machine learning optimizes ad delivery
TikTok's delivery system learns from every impression and click. It matches your ad to users with the highest probability of converting. But it needs scale to learn. Narrow audiences reduce that scale and slow the learning phase down significantly.
Think of targeting as guardrails, not a funnel. Set the outer boundaries, then let TikTok fill in the detail.
Key metrics: CPA, conversion rate, and ROAS
Three metrics drive targeting decisions: cost-per-acquisition (CPA), conversion rate, and return on ad spend (ROAS). CPA tells you what each customer costs. Conversion rate shows how well your audience-to-landing-page match is working. ROAS confirms the spend is worth it.
Track all three together. A falling CPA alongside falling ROAS means you're buying cheap but unqualified conversions. Balance matters.
---
Five Optimization Strategies That Work
These five strategies reflect TikTok Ads Manager's own guidance, combined with proven advertiser patterns.
Start with broad targeting (and test before narrowing)
Launch your first campaigns with minimal restrictions. No interests, no behaviors, basic demographics only if legally required. Let TikTok's algorithm gather data for at least one week.
After seven days, review your CPA trends. If broad targeting performs well, don't layer more restrictions out of habit. Restrictions should earn their place with evidence.
Enable Smart Targeting toggles for performance stability
TikTok's Smart Targeting includes two toggles: smart audience and smart interests and behavior. Enabling both gives TikTok permission to expand delivery beyond your manual settings when it predicts better results.
Smart Targeting improves CPA by an average of 5% for web conversion advertisers, per TikTok's documentation. One important note: Smart Targeting will not expand into excluded audiences or demographic variables like age or gender. Your exclusions stay firm. Also note Smart Targeting is not available for Global App Bundle or Pangle placements.
Build and leverage custom audiences for retargeting
Custom audiences let you re-engage people who already know your brand. Website visitors, video viewers, app users, past engagers. These audiences convert at higher rates because the trust barrier is lower.
Per TikTok Ads Manager documentation, custom audiences also let you exclude recent customers, protecting spend for prospecting campaigns. The top 50% of TikTok's highest-spending advertisers incorporate custom audiences into over 50% of their ad spend. That's not coincidence.
Create lookalike audiences for prospecting at scale
Once your custom audience hits 1,000 users (the minimum required), you can build a lookalike. TikTok finds users with similar signals to your best customers.
Lookalikes are your fastest path to cold-audience scale without starting from zero. Start with a 1-5% similarity range for tighter match quality. Expand to 5-10% when you need more volume.
Use manual controls strategically (geography, spending power)
Geography and spending power are the two manual controls worth keeping. They're based on hard facts about your customer base, not guesses about interests.
If you only ship to the US, restrict to the US. If your product price point requires disposable income, layer in spending power filters. Everything else, let the algorithm decide.
---
Testing and Validation Framework
Systematic testing separates advertisers who improve from those who guess.
A/B test targeting variables over 7+ days
TikTok offers three split-test types: targeting, bidding and optimization, and creative. For targeting tests, run each variant for at least seven days before drawing conclusions. Shorter windows produce noisy data.
Change one variable per test. Testing geography and audience type at the same time makes it impossible to know what worked.
Monitor CPA, conversion rate, and ROAS trends
Check your metrics daily but make decisions weekly. Daily swings are normal. Weekly trends reveal real patterns.
Set up a simple tracking sheet. Record CPA, conversion rate, and ROAS for each ad set every seven days. Look for directional movement, not single-day spikes.
Always validate advanced targeting against broad baseline
Before scaling any advanced targeting setup, pit it against a broad baseline in a proper A/B test. TikTok's own guidance recommends this step. It's easy to assume custom audiences or lookalikes will outperform broad. Sometimes they do. Sometimes they don't. Let the data confirm before you shift budget.
Set realistic improvement targets (10-20% CPA reduction)
Target 10-20% CPA reduction per quarter. That pace is realistic and compounds well. If a targeting change appears to cut CPA by 50% overnight, investigate before celebrating. Sudden dramatic shifts usually reflect a measurement issue, not a breakthrough. Steady improvement beats volatile wins.
---
How Coinis Powers Targeting Success
Targeting optimization only works when your creative matches the audience. The best targeting setup fails with the wrong ad.
Generate and test multiple creative variations faster
Coinis generates ad images from a product URL using premium AI models. You can spin up multiple creative variants in minutes. More variants mean more signal across your targeting tests.
Direct publishing from Coinis to TikTok is on the roadmap. Today, you export creatives and upload them to TikTok Ads Manager directly. The creative-building speed still cuts your production time significantly.
Use Revise to adapt creatives for audience segments
Different audience segments respond to different messages. Coinis Revise lets you adapt creatives without rebuilding from scratch. Change text on an image, swap colors, erase an object, or translate copy to another language. Each edit takes one click.
If you're targeting a Spanish-speaking segment separately from an English-speaking one, AI Translate handles the switch in seconds.
Pair AI copywriting with advanced audience insights
Coinis AI Copywriting generates headlines, body copy, and CTAs from your Brand Profile. When you know what makes a segment tick (from TikTok's audience insights), you can brief the AI accordingly. The Brand Profile stores your voice so copy stays consistent across segments.
Track performance and refine creative alongside targeting
Coinis's Advertise page delivers live performance reporting on Meta campaigns today. For TikTok, pull data from TikTok Ads Manager and use those insights to decide which Coinis-generated creatives to iterate next. Targeting and creative decisions reinforce each other. Improve both together.
---
Key Takeaways
- Broad targeting (80%+ reach) delivers 15% lower CPA and 20% higher conversion rate than narrow setups.
- Smart Targeting toggles add an average 5% CPA improvement for web conversion advertisers.
- Custom audiences and lookalikes scale best once your pixel has at least 1,000 users to seed from.
- Run every targeting test for at least 7 days. Make weekly decisions, not daily ones.
- Aim for 10-20% CPA improvement per quarter. Sudden dramatic gains usually signal a measurement error.
- Pair every targeting change with a creative refresh. Audience and ad must match.
---
Or let Coinis do it.
From a product URL to a live Meta campaign. AI-generated creatives. On-brand copy. Direct publish to Facebook and Instagram. Real performance reporting. All in one platform.
Start free. Upgrade when you're ready.
15 AI tokens a month. No credit card.
Frequently Asked Questions
Should I use broad or narrow targeting for TikTok ads?
Start broad. TikTok's internal testing shows broad targeting (80%+ audience reach) achieves 15% lower CPA and 20% higher conversion rate compared to narrow targeting. Let the algorithm find buyers. Layer restrictions only when data tells you to.
What is Smart Targeting on TikTok and should I enable it?
Smart Targeting includes two toggles in TikTok Ads Manager: smart audience and smart interests and behavior. Enabling them lets TikTok expand delivery when it predicts better results. Per TikTok's documentation, it improves CPA by an average of 5% for web conversion advertisers. Note: Smart Targeting won't expand into your excluded audiences or demographic filters (age, gender), and it's not available for Pangle or Global App Bundle placements.
How long should I run a TikTok targeting A/B test?
At least 7 days. Shorter test windows produce noisy, unreliable data. TikTok's machine learning needs time to optimize delivery, so early results often don't reflect final performance. Make decisions on weekly trends, not daily swings.
How do lookalike audiences work on TikTok?
You build a lookalike from a custom audience with a minimum of 1,000 users. TikTok finds users with similar signals to that seed audience. Start with a 1-5% similarity range for quality, and expand to 5-10% when you need more volume.